[
{"id":"lit_2405_08644","type":"article","title":"Thinking Tokens for Language Modeling","author":[{"literal":"David Herel"},{"literal":"Tomas Mikolov"}],"issued":{"date-parts":[[2024,5,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2405.08644","URL":"https://arxiv.org/abs/2405.08644","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2405.08644/","first_public":"2023-09-04","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2310_02226","type":"article","title":"Think before you speak: Training language models with pause tokens","author":[{"literal":"Sachin Goyal"},{"literal":"Ziwei Ji"},{"literal":"Ankit Singh Rawat"},{"literal":"Aditya Krishna Menon"},{"literal":"Sanjiv Kumar"},{"literal":"Vaishnavh Nagarajan"}],"issued":{"date-parts":[[2023,10,3]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2310.02226","URL":"https://arxiv.org/abs/2310.02226","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2310.02226/","first_public":"2023-10-03","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2310_05707","type":"article","title":"Guiding Language Model Reasoning with Planning Tokens","author":[{"literal":"Xinyi Wang"},{"literal":"Lucas Caccia"},{"literal":"Oleksiy Ostapenko"},{"literal":"Xingdi Yuan"},{"literal":"William Yang Wang"},{"literal":"Alessandro Sordoni"}],"issued":{"date-parts":[[2023,10,9]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2310.05707","URL":"https://arxiv.org/abs/2310.05707","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2310.05707/","first_public":"2023-10-09","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2310_10845","type":"article","title":"CoTFormer: A Chain-of-Thought Driven Architecture with Budget-Adaptive Computation Cost at Inference","author":[{"literal":"Amirkeivan Mohtashami"},{"literal":"Matteo Pagliardini"},{"literal":"Martin Jaggi"}],"issued":{"date-parts":[[2023,10,16]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2310.10845","URL":"https://arxiv.org/abs/2310.10845","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2310.10845/","first_public":"2023-10-16","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2310_14491","type":"article","title":"Towards a Mechanistic Interpretation of Multi-Step Reasoning Capabilities of Language Models","author":[{"literal":"Yifan Hou"},{"literal":"Jiaoda Li"},{"literal":"Yu Fei"},{"literal":"Alessandro Stolfo"},{"literal":"Wangchunshu Zhou"},{"literal":"Guangtao Zeng"},{"literal":"Antoine Bosselut"},{"literal":"Mrinmaya Sachan"}],"issued":{"date-parts":[[2023,10,23]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2310.14491","URL":"https://arxiv.org/abs/2310.14491","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2310.14491/","first_public":"2023-10-23","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2311_01460","type":"article","title":"Implicit Chain of Thought Reasoning via Knowledge Distillation","author":[{"literal":"Yuntian Deng"},{"literal":"Kiran Prasad"},{"literal":"Roland Fernandez"},{"literal":"Paul Smolensky"},{"literal":"Vishrav Chaudhary"},{"literal":"Stuart Shieber"}],"issued":{"date-parts":[[2023,11,2]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2311.01460","URL":"https://arxiv.org/abs/2311.01460","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2311.01460/","first_public":"2023-11-02","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2402_07754","type":"article","title":"Diffusion of Thoughts: Chain-of-Thought Reasoning in Diffusion Language Models","author":[{"literal":"Jiacheng Ye"},{"literal":"Shansan Gong"},{"literal":"Liheng Chen"},{"literal":"Lin Zheng"},{"literal":"Jiahui Gao"},{"literal":"Han Shi"},{"literal":"Chuan Wu"},{"literal":"Xin Jiang"},{"literal":"Zhenguo Li"},{"literal":"Wei Bi"},{"literal":"Lingpeng Kong"}],"issued":{"date-parts":[[2024,2,12]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2402.07754","URL":"https://arxiv.org/abs/2402.07754","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2402.07754/","first_public":"2024-02-12","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2402_11917","type":"article","title":"A Mechanistic Analysis of a Transformer Trained on a Symbolic Multi-Step Reasoning Task","author":[{"literal":"Jannik Brinkmann"},{"literal":"Abhay Sheshadri"},{"literal":"Victor Levoso"},{"literal":"Paul Swoboda"},{"literal":"Christian Bartelt"}],"issued":{"date-parts":[[2024,2,19]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2402.11917","URL":"https://arxiv.org/abs/2402.11917","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2402.11917/","first_public":"2024-02-19","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2402_14328","type":"article","title":"Understanding and Patching Compositional Reasoning in LLMs","author":[{"literal":"Zhaoyi Li"},{"literal":"Gangwei Jiang"},{"literal":"Hong Xie"},{"literal":"Linqi Song"},{"literal":"Defu Lian"},{"literal":"Ying Wei"}],"issued":{"date-parts":[[2024,2,22]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2402.14328","URL":"https://arxiv.org/abs/2402.14328","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2402.14328/","first_public":"2024-02-22","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2402_16837","type":"article","title":"Do Large Language Models Latently Perform Multi-Hop Reasoning?","author":[{"literal":"Sohee Yang"},{"literal":"Elena Gribovskaya"},{"literal":"Nora Kassner"},{"literal":"Mor Geva"},{"literal":"Sebastian Riedel"}],"issued":{"date-parts":[[2024,2,26]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2402.16837","URL":"https://arxiv.org/abs/2402.16837","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2402.16837/","first_public":"2024-02-26","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2403_09629","type":"article","title":"Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking","author":[{"literal":"Eric Zelikman"},{"literal":"Georges Harik"},{"literal":"Yijia Shao"},{"literal":"Varuna Jayasiri"},{"literal":"Nick Haber"},{"literal":"Noah D. Goodman"}],"issued":{"date-parts":[[2024,3,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2403.09629","URL":"https://arxiv.org/abs/2403.09629","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2403.09629/","first_public":"2024-03-14","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2404_15758","type":"article","title":"Let's think dot by dot: Hidden computation in transformer language models","author":[{"literal":"Jacob Pfau"},{"literal":"William Merrill"},{"literal":"Samuel R. Bowman"}],"issued":{"date-parts":[[2024,4,24]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2404.15758","URL":"https://arxiv.org/abs/2404.15758","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2404.15758/","first_public":"2024-04-24","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2405_14838","type":"article","title":"From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step","author":[{"literal":"Yuntian Deng"},{"literal":"Yejin Choi"},{"literal":"Stuart Shieber"}],"issued":{"date-parts":[[2024,5,23]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2405.14838","URL":"https://arxiv.org/abs/2405.14838","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2405.14838/","first_public":"2024-05-23","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2405_15071","type":"article","title":"Grokked Transformers are Implicit Reasoners: A Mechanistic Journey to the Edge of Generalization","author":[{"literal":"Boshi Wang"},{"literal":"Xiang Yue"},{"literal":"Yu Su"},{"literal":"Huan Sun"}],"issued":{"date-parts":[[2024,5,23]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2405.15071","URL":"https://arxiv.org/abs/2405.15071","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2405.15071/","first_public":"2024-05-23","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2406_13858","type":"article","title":"Distributional Reasoning in LLMs: Parallel Reasoning Processes in Multi-Hop Reasoning","author":[{"literal":"Yuval Shalev"},{"literal":"Amir Feder"},{"literal":"Ariel Goldstein"}],"issued":{"date-parts":[[2024,6,19]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2406.13858","URL":"https://arxiv.org/abs/2406.13858","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2406.13858/","first_public":"2024-06-19","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2407_06023","type":"article","title":"Distilling System 2 into System 1","author":[{"literal":"Ping Yu"},{"literal":"Jing Xu"},{"literal":"Jason Weston"},{"literal":"Ilia Kulikov"}],"issued":{"date-parts":[[2024,7,8]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2407.06023","URL":"https://arxiv.org/abs/2407.06023","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2407.06023/","first_public":"2024-07-08","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2407_20311","type":"article","title":"Physics of Language Models: Part 2.1, Grade-School Math and the Hidden Reasoning Process","author":[{"literal":"Tian Ye"},{"literal":"Zicheng Xu"},{"literal":"Yuanzhi Li"},{"literal":"Zeyuan Allen-Zhu"}],"issued":{"date-parts":[[2024,7,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2407.20311","URL":"https://arxiv.org/abs/2407.20311","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2407.20311/","first_public":"2024-07-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2409_08561","type":"article","title":"Expediting and Elevating Large Language Model Reasoning via Hidden Chain-of-Thought Decoding","author":[{"literal":"Tianqiao Liu"},{"literal":"Zui Chen"},{"literal":"Zitao Liu"},{"literal":"Mi Tian"},{"literal":"Weiqi Luo"}],"issued":{"date-parts":[[2024,9,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2409.08561","URL":"https://arxiv.org/abs/2409.08561","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2409.08561/","first_public":"2024-09-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2409_14026","type":"article","title":"Uncovering Latent Chain of Thought Vectors in Language Models","author":[{"literal":"Jason Zhang"},{"literal":"Scott Viteri"}],"issued":{"date-parts":[[2024,9,21]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2409.14026","URL":"https://arxiv.org/abs/2409.14026","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2409.14026/","first_public":"2024-09-21","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2410_09918","type":"article","title":"Dualformer: Controllable Fast and Slow Thinking by Learning with Randomized Reasoning Traces","author":[{"literal":"DiJia Su"},{"literal":"Sainbayar Sukhbaatar"},{"literal":"Michael Rabbat"},{"literal":"Yuandong Tian"},{"literal":"Qinqing Zheng"}],"issued":{"date-parts":[[2024,10,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2410.09918","URL":"https://arxiv.org/abs/2410.09918","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2410.09918/","first_public":"2024-10-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2410_13640","type":"article","title":"Latent Space Chain-of-Embedding Enables Output-free LLM Self-Evaluation","author":[{"literal":"Yiming Wang"},{"literal":"Pei Zhang"},{"literal":"Baosong Yang"},{"literal":"Derek F. Wong"},{"literal":"Rui Wang"}],"issued":{"date-parts":[[2024,10,17]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2410.13640","URL":"https://arxiv.org/abs/2410.13640","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2410.13640/","first_public":"2024-10-17","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2411_15862","type":"article","title":"Do LLMs Really Think Step-by-step In Implicit Reasoning?","author":[{"literal":"Yijiong Yu"}],"issued":{"date-parts":[[2024,11,24]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2411.15862","URL":"https://arxiv.org/abs/2411.15862","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2411.15862/","first_public":"2024-11-24","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2412_06769","type":"article","title":"Training Large Language Models to Reason in a Continuous Latent Space","author":[{"literal":"Shibo Hao"},{"literal":"Sainbayar Sukhbaatar"},{"literal":"DiJia Su"},{"literal":"Xian Li"},{"literal":"Zhiting Hu"},{"literal":"Jason Weston"},{"literal":"Yuandong Tian"}],"issued":{"date-parts":[[2024,12,9]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2412.06769","URL":"https://arxiv.org/abs/2412.06769","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2412.06769/","first_public":"2024-12-09","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2412_08635","type":"article","title":"Multimodal Latent Language Modeling with Next-Token Diffusion","author":[{"literal":"Yutao Sun"},{"literal":"Hangbo Bao"},{"literal":"Wenhui Wang"},{"literal":"Zhiliang Peng"},{"literal":"Li Dong"},{"literal":"Shaohan Huang"},{"literal":"Jianyong Wang"},{"literal":"Furu Wei"}],"issued":{"date-parts":[[2024,12,11]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2412.08635","URL":"https://arxiv.org/abs/2412.08635","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2412.08635/","first_public":"2024-12-11","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2412_13171","type":"article","title":"Compressed Chain of Thought: Efficient Reasoning Through Dense Representations","author":[{"literal":"Jeffrey Cheng"},{"literal":"Benjamin Van Durme"}],"issued":{"date-parts":[[2024,12,17]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2412.13171","URL":"https://arxiv.org/abs/2412.13171","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2412.13171/","first_public":"2024-12-17","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2501_19201","type":"article","title":"Efficient Reasoning with Hidden Thinking","author":[{"literal":"Xuan Shen"},{"literal":"Yizhou Wang"},{"literal":"Yufa Zhou"},{"literal":"Xiangxi Shi"},{"literal":"Pu Zhao"},{"literal":"Yanzhi Wang"},{"literal":"Jiuxiang Gu"}],"issued":{"date-parts":[[2025,1,31]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2501.19201","URL":"https://arxiv.org/abs/2501.19201","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2501.19201/","first_public":"2025-01-31","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2502_03275","type":"article","title":"Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning","author":[{"literal":"DiJia Su"},{"literal":"Hanlin Zhu"},{"literal":"Yingchen Xu"},{"literal":"Jiantao Jiao"},{"literal":"Yuandong Tian"},{"literal":"Qinqing Zheng"}],"issued":{"date-parts":[[2025,2,5]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2502.03275","URL":"https://arxiv.org/abs/2502.03275","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2502.03275/","first_public":"2025-02-05","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2502_05171","type":"article","title":"Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach","author":[{"literal":"Jonas Geiping"},{"literal":"Sean McLeish"},{"literal":"Neel Jain"},{"literal":"John Kirchenbauer"},{"literal":"Siddharth Singh"},{"literal":"Brian R. Bartoldson"},{"literal":"Bhavya Kailkhura"},{"literal":"Abhinav Bhatele"},{"literal":"Tom Goldstein"}],"issued":{"date-parts":[[2025,2,7]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2502.05171","URL":"https://arxiv.org/abs/2502.05171","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2502.05171/","first_public":"2025-02-07","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2502_07827","type":"article","title":"Implicit Language Models are RNNs: Balancing Parallelization and Expressivity","author":[{"literal":"Mark Schöne"},{"literal":"Babak Rahmani"},{"literal":"Heiner Kremer"},{"literal":"Fabian Falck"},{"literal":"Hitesh Ballani"},{"literal":"Jannes Gladrow"}],"issued":{"date-parts":[[2025,2,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2502.07827","URL":"https://arxiv.org/abs/2502.07827","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2502.07827/","first_public":"2025-02-10","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2502_08524","type":"article","title":"LLM Pretraining with Continuous Concepts","author":[{"literal":"Jihoon Tack"},{"literal":"Jack Lanchantin"},{"literal":"Jane Yu"},{"literal":"Andrew Cohen"},{"literal":"Ilia Kulikov"},{"literal":"Janice Lan"},{"literal":"Shibo Hao"},{"literal":"Yuandong Tian"},{"literal":"Jason Weston"},{"literal":"Xian Li"}],"issued":{"date-parts":[[2025,2,12]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2502.08524","URL":"https://arxiv.org/abs/2502.08524","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2502.08524/","first_public":"2025-02-12","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2502_13842","type":"article","title":"Inner Thinking Transformer: Leveraging Dynamic Depth Scaling to Foster Adaptive Internal Thinking","author":[{"literal":"Yilong Chen"},{"literal":"Junyuan Shang"},{"literal":"Zhenyu Zhang"},{"literal":"Yanxi Xie"},{"literal":"Jiawei Sheng"},{"literal":"Tingwen Liu"},{"literal":"Shuohuan Wang"},{"literal":"Yu Sun"},{"literal":"Hua Wu"},{"literal":"Haifeng Wang"}],"issued":{"date-parts":[[2025,2,19]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2502.13842","URL":"https://arxiv.org/abs/2502.13842","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2502.13842/","first_public":"2025-02-19","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2502_15589","type":"article","title":"LightThinker: Thinking Step-by-Step Compression","author":[{"literal":"Jintian Zhang"},{"literal":"Yuqi Zhu"},{"literal":"Mengshu Sun"},{"literal":"Yujie Luo"},{"literal":"Shuofei Qiao"},{"literal":"Lun Du"},{"literal":"Da Zheng"},{"literal":"Huajun Chen"},{"literal":"Ningyu Zhang"}],"issued":{"date-parts":[[2025,2,21]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2502.15589","URL":"https://arxiv.org/abs/2502.15589","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2502.15589/","first_public":"2025-02-21","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2502_17416","type":"article","title":"Reasoning with Latent Thoughts: On the Power of Looped Transformers","author":[{"literal":"Nikunj Saunshi"},{"literal":"Nishanth Dikkala"},{"literal":"Zhiyuan Li"},{"literal":"Sanjiv Kumar"},{"literal":"Sashank J. Reddi"}],"issued":{"date-parts":[[2025,2,24]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2502.17416","URL":"https://arxiv.org/abs/2502.17416","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2502.17416/","first_public":"2025-02-24","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2502_21074","type":"article","title":"CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation","author":[{"literal":"Zhenyi Shen"},{"literal":"Hanqi Yan"},{"literal":"Linhai Zhang"},{"literal":"Zhanghao Hu"},{"literal":"Yali Du"},{"literal":"Yulan He"}],"issued":{"date-parts":[[2025,2,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2502.21074","URL":"https://arxiv.org/abs/2502.21074","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2502.21074/","first_public":"2025-02-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2503_07604","type":"article","title":"Implicit Reasoning in Transformers is Reasoning through Shortcuts","author":[{"literal":"Tianhe Lin"},{"literal":"Jian Xie"},{"literal":"Siyu Yuan"},{"literal":"Deqing Yang"}],"issued":{"date-parts":[[2025,3,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2503.07604","URL":"https://arxiv.org/abs/2503.07604","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2503.07604/","first_public":"2025-03-10","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2504_10615","type":"article","title":"Beyond Chains of Thought: Benchmarking Latent-Space Reasoning Abilities in Large Language Models","author":[{"literal":"Thilo Hagendorff"},{"literal":"Sarah Fabi"}],"issued":{"date-parts":[[2025,4,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2504.10615","URL":"https://arxiv.org/abs/2504.10615","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2504.10615/","first_public":"2025-04-14","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2504_14992","type":"article","title":"Efficient Pretraining Length Scaling","author":[{"literal":"Bohong Wu"},{"literal":"Shen Yan"},{"literal":"Sijun Zhang"},{"literal":"Jianqiao Lu"},{"literal":"Yutao Zeng"},{"literal":"Ya Wang"},{"literal":"Xun Zhou"}],"issued":{"date-parts":[[2025,4,21]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2504.14992","URL":"https://arxiv.org/abs/2504.14992","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2504.14992/","first_public":"2025-04-21","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_12629","type":"article","title":"Enhancing Latent Computation in Transformers with Latent Tokens","author":[{"literal":"Yuchang Sun"},{"literal":"Yanxi Chen"},{"literal":"Yaliang Li"},{"literal":"Bolin Ding"}],"issued":{"date-parts":[[2025,5,19]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.12629","URL":"https://arxiv.org/abs/2505.12629","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.12629/","first_public":"2025-05-19","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_13308","type":"article","title":"Seek in the Dark: Reasoning via Test-Time Instance-Level Policy Gradient in Latent Space","author":[{"literal":"Hengli Li"},{"literal":"Chenxi Li"},{"literal":"Tong Wu"},{"literal":"Xuekai Zhu"},{"literal":"Yuxuan Wang"},{"literal":"Zhaoxin Yu"},{"literal":"Eric Hanchen Jiang"},{"literal":"Song-Chun Zhu"},{"literal":"Zixia Jia"},{"literal":"Ying Nian Wu"},{"literal":"Zilong Zheng"}],"issued":{"date-parts":[[2025,5,19]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.13308","URL":"https://arxiv.org/abs/2505.13308","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.13308/","first_public":"2025-05-19","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_13898","type":"article","title":"Do Language Models Use Their Depth Efficiently?","author":[{"literal":"Róbert Csordás"},{"literal":"Christopher D. Manning"},{"literal":"Christopher Potts"}],"issued":{"date-parts":[[2025,5,20]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.13898","URL":"https://arxiv.org/abs/2505.13898","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.13898/","first_public":"2025-05-20","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_14530","type":"article","title":"Internal Chain-of-Thought: Empirical Evidence for Layer-wise Subtask Scheduling in LLMs","author":[{"literal":"Zhipeng Yang"},{"literal":"Junzhuo Li"},{"literal":"Siyu Xia"},{"literal":"Xuming Hu"}],"issued":{"date-parts":[[2025,5,20]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.14530","URL":"https://arxiv.org/abs/2505.14530","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.14530/","first_public":"2025-05-20","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_14827","type":"article","title":"Text Generation Beyond Discrete Token Sampling","author":[{"literal":"Yufan Zhuang"},{"literal":"Liyuan Liu"},{"literal":"Chandan Singh"},{"literal":"Jingbo Shang"},{"literal":"Jianfeng Gao"}],"issued":{"date-parts":[[2025,5,20]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.14827","URL":"https://arxiv.org/abs/2505.14827","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.14827/","first_public":"2025-05-20","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_16552","type":"article","title":"Think Silently, Think Fast: Dynamic Latent Compression of LLM Reasoning Chains","author":[{"literal":"Wenhui Tan"},{"literal":"Jiaze Li"},{"literal":"Jianzhong Ju"},{"literal":"Zhenbo Luo"},{"literal":"Ruihua Song"},{"literal":"Jian Luan"}],"issued":{"date-parts":[[2025,5,22]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.16552","URL":"https://arxiv.org/abs/2505.16552","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.16552/","first_public":"2025-05-22","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_16782","type":"article","title":"Reasoning Beyond Language: A Comprehensive Survey on Latent Chain-of-Thought Reasoning","author":[{"literal":"Xinghao Chen"},{"literal":"Anhao Zhao"},{"literal":"Heming Xia"},{"literal":"Xuan Lu"},{"literal":"Hanlin Wang"},{"literal":"Yanjun Chen"},{"literal":"Wei Zhang"},{"literal":"Jian Wang"},{"literal":"Wenjie Li"},{"literal":"Xiaoyu Shen"}],"issued":{"date-parts":[[2025,5,22]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.16782","URL":"https://arxiv.org/abs/2505.16782","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.16782/","first_public":"2025-05-22","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_17923","type":"article","title":"Language Models Can Learn Implicit Multi-Hop Reasoning, but Only If They Have Lots of Training Data","author":[{"literal":"Yuekun Yao"},{"literal":"Yupei Du"},{"literal":"Dawei Zhu"},{"literal":"Michael Hahn"},{"literal":"Alexander Koller"}],"issued":{"date-parts":[[2025,5,23]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.17923","URL":"https://arxiv.org/abs/2505.17923","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.17923/","first_public":"2025-05-23","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_18454","type":"article","title":"Hybrid Latent Reasoning via Reinforcement Learning","author":[{"literal":"Zhenrui Yue"},{"literal":"Bowen Jin"},{"literal":"Huimin Zeng"},{"literal":"Honglei Zhuang"},{"literal":"Zhen Qin"},{"literal":"Jinsung Yoon"},{"literal":"Lanyu Shang"},{"literal":"Jiawei Han"},{"literal":"Dong Wang"}],"issued":{"date-parts":[[2025,5,24]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.18454","URL":"https://arxiv.org/abs/2505.18454","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.18454/","first_public":"2025-05-24","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_18962","type":"article","title":"System-1.5 Reasoning: Traversal in Language and Latent Spaces with Dynamic Shortcuts","author":[{"literal":"Xiaoqiang Wang"},{"literal":"Suyuchen Wang"},{"literal":"Yun Zhu"},{"literal":"Bang Liu"}],"issued":{"date-parts":[[2025,5,25]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.18962","URL":"https://arxiv.org/abs/2505.18962","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.18962/","first_public":"2025-05-25","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_19245","type":"article","title":"To CoT or To Loop? A Formal Comparison Between Chain-of-Thought and Looped Transformers","author":[{"literal":"Kevin Xu"},{"literal":"Issei Sato"}],"issued":{"date-parts":[[2025,5,25]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.19245","URL":"https://arxiv.org/abs/2505.19245","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.19245/","first_public":"2025-05-25","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_20674","type":"article","title":"PonderLM: Pretraining Language Models to Ponder in Continuous Space","author":[{"literal":"Boyi Zeng"},{"literal":"Shixiang Song"},{"literal":"Siyuan Huang"},{"literal":"Yixuan Wang"},{"literal":"He Li"},{"literal":"Ziwei He"},{"literal":"Xinbing Wang"},{"literal":"Zhiyu Li"},{"literal":"Zhouhan Lin"}],"issued":{"date-parts":[[2025,5,27]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.20674","URL":"https://arxiv.org/abs/2505.20674","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.20674/","first_public":"2025-05-27","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2505_23648","type":"article","title":"Continuous Chain of Thought Enables Parallel Exploration and Reasoning","author":[{"literal":"Halil Alperen Gozeten"},{"literal":"M. Emrullah Ildiz"},{"literal":"Xuechen Zhang"},{"literal":"Hrayr Harutyunyan"},{"literal":"Ankit Singh Rawat"},{"literal":"Samet Oymak"}],"issued":{"date-parts":[[2025,5,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2505.23648","URL":"https://arxiv.org/abs/2505.23648","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2505.23648/","first_public":"2025-05-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2506_11752","type":"article","title":"DART: Distilling Autoregressive Reasoning to Silent Thought","author":[{"literal":"Nan Jiang"},{"literal":"Ziming Wu"},{"literal":"De-Chuan Zhan"},{"literal":"Fuming Lai"},{"literal":"Shaobing Lian"}],"issued":{"date-parts":[[2025,6,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2506.11752","URL":"https://arxiv.org/abs/2506.11752","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2506.11752/","first_public":"2025-06-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2506_17218","type":"article","title":"Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens","author":[{"literal":"Zeyuan Yang"},{"literal":"Xueyang Yu"},{"literal":"Delin Chen"},{"literal":"Maohao Shen"},{"literal":"Chuang Gan"}],"issued":{"date-parts":[[2025,6,20]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2506.17218","URL":"https://arxiv.org/abs/2506.17218","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2506.17218/","first_public":"2025-06-20","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2506_18582","type":"article","title":"Parallel Continuous Chain-of-Thought with Jacobi Iteration","author":[{"literal":"Haoyi Wu"},{"literal":"Zhihao Teng"},{"literal":"Kewei Tu"}],"issued":{"date-parts":[[2025,6,23]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2506.18582","URL":"https://arxiv.org/abs/2506.18582","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2506.18582/","first_public":"2025-06-23","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2506_21734","type":"article","title":"Hierarchical Reasoning Model","author":[{"literal":"Guan Wang"},{"literal":"Jin Li"},{"literal":"Yuhao Sun"},{"literal":"Xing Chen"},{"literal":"Changling Liu"},{"literal":"Yue Wu"},{"literal":"Meng Lu"},{"literal":"Sen Song"},{"literal":"Yasin Abbasi Yadkori"}],"issued":{"date-parts":[[2025,6,26]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2506.21734","URL":"https://arxiv.org/abs/2506.21734","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2506.21734/","first_public":"2025-06-26","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2507_02199","type":"article","title":"Latent Chain-of-Thought? Decoding the Depth-Recurrent Transformer","author":[{"literal":"Wenquan Lu"},{"literal":"Yuechuan Yang"},{"literal":"Kyle Lee"},{"literal":"Yanshu Li"},{"literal":"Enqi Liu"}],"issued":{"date-parts":[[2025,7,2]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2507.02199","URL":"https://arxiv.org/abs/2507.02199","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2507.02199/","first_public":"2025-07-02","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2507_06203","type":"article","title":"A Survey on Latent Reasoning","author":[{"literal":"Rui-Jie Zhu"},{"literal":"Tianhao Peng"},{"literal":"Tianhao Cheng"},{"literal":"Xingwei Qu"},{"literal":"Jinfa Huang"},{"literal":"Dawei Zhu"},{"literal":"Hao Wang"},{"literal":"Kaiwen Xue"},{"literal":"Xuanliang Zhang"},{"literal":"Yong Shan"},{"literal":"Tianle Cai"},{"literal":"Taylor Kergan"},{"literal":"Assel Kembay"},{"literal":"Andrew Smith"},{"literal":"Chenghua Lin"},{"literal":"Binh Nguyen"},{"literal":"Yuqi Pan"},{"literal":"Yuhong Chou"},{"literal":"Zefan Cai"},{"literal":"Zhenhe Wu"},{"literal":"Yongchi Zhao"},{"literal":"Tianyu Liu"},{"literal":"Jian Yang"},{"literal":"Wangchunshu Zhou"},{"literal":"Chujie Zheng"},{"literal":"Chongxuan Li"},{"literal":"Yuyin Zhou"},{"literal":"Zhoujun Li"},{"literal":"Zhaoxiang Zhang"},{"literal":"Jiaheng Liu"},{"literal":"Ge Zhang"},{"literal":"Wenhao Huang"},{"literal":"Jason Eshraghian"}],"issued":{"date-parts":[[2025,7,8]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2507.06203","URL":"https://arxiv.org/abs/2507.06203","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2507.06203/","first_public":"2025-07-08","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2507_07996","type":"article","title":"Skip a Layer or Loop it? Test-Time Depth Adaptation of Pretrained LLMs","author":[{"literal":"Ziyue Li"},{"literal":"Yang Li"},{"literal":"Tianyi Zhou"}],"issued":{"date-parts":[[2025,7,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2507.07996","URL":"https://arxiv.org/abs/2507.07996","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2507.07996/","first_public":"2025-07-10","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2507_10524","type":"article","title":"Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computation","author":[{"literal":"Sangmin Bae"},{"literal":"Yujin Kim"},{"literal":"Reza Bayat"},{"literal":"Sungnyun Kim"},{"literal":"Jiyoun Ha"},{"literal":"Tal Schuster"},{"literal":"Adam Fisch"},{"literal":"Hrayr Harutyunyan"},{"literal":"Ziwei Ji"},{"literal":"Aaron Courville"},{"literal":"Se-Young Yun"}],"issued":{"date-parts":[[2025,7,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2507.10524","URL":"https://arxiv.org/abs/2507.10524","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2507.10524/","first_public":"2025-07-14","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2508_00574","type":"article","title":"SynAdapt: Learning Adaptive Reasoning in Large Language Models via Synthetic Continuous Chain-of-Thought","author":[{"literal":"Jianwei Wang"},{"literal":"Ziming Wu"},{"literal":"Fuming Lai"},{"literal":"Shaobing Lian"},{"literal":"Ziqian Zeng"}],"issued":{"date-parts":[[2025,8,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2508.00574","URL":"https://arxiv.org/abs/2508.00574","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2508.00574/","first_public":"2025-08-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2508_03440","type":"article","title":"LLMs are Single-threaded Reasoners: Demystifying the Working Mechanism of Soft Thinking","author":[{"literal":"Junhong Wu"},{"literal":"Jinliang Lu"},{"literal":"Zixuan Ren"},{"literal":"Gangqiang Hu"},{"literal":"Zhi Wu"},{"literal":"Dai Dai"},{"literal":"Hua Wu"}],"issued":{"date-parts":[[2025,8,5]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2508.03440","URL":"https://arxiv.org/abs/2508.03440","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2508.03440/","first_public":"2025-08-05","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2509_02350","type":"article","title":"Implicit Reasoning in Large Language Models: A Comprehensive Survey","author":[{"literal":"Jindong Li"},{"literal":"Yali Fu"},{"literal":"Li Fan"},{"literal":"Jiahong Liu"},{"literal":"Yao Shu"},{"literal":"Chengwei Qin"},{"literal":"Menglin Yang"},{"literal":"Irwin King"},{"literal":"Rex Ying"}],"issued":{"date-parts":[[2025,9,2]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2509.02350","URL":"https://arxiv.org/abs/2509.02350","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2509.02350/","first_public":"2025-09-02","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2509_20317","type":"article","title":"SIM-CoT: Supervised Implicit Chain-of-Thought","author":[{"literal":"Xilin Wei"},{"literal":"Xiaoran Liu"},{"literal":"Yuhang Zang"},{"literal":"Xiaoyi Dong"},{"literal":"Yuhang Cao"},{"literal":"Jiaqi Wang"},{"literal":"Xipeng Qiu"},{"literal":"Dahua Lin"}],"issued":{"date-parts":[[2025,9,24]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2509.20317","URL":"https://arxiv.org/abs/2509.20317","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2509.20317/","first_public":"2025-09-24","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2510_04573","type":"article","title":"LaDiR: Latent Diffusion Enhances LLMs for Text Reasoning","author":[{"literal":"Haoqiang Kang"},{"literal":"Yizhe Zhang"},{"literal":"Nikki Lijing Kuang"},{"literal":"Nicklas Majamaki"},{"literal":"Navdeep Jaitly"},{"literal":"Yi-An Ma"},{"literal":"Lianhui Qin"}],"issued":{"date-parts":[[2025,10,6]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2510.04573","URL":"https://arxiv.org/abs/2510.04573","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2510.04573/","first_public":"2025-10-06","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2510_15522","type":"article","title":"LLM Latent Reasoning as Chain of Superposition","author":[{"literal":"Jingcheng Deng"},{"literal":"Liang Pang"},{"literal":"Zihao Wei"},{"literal":"Shicheng Xu"},{"literal":"Zenghao Duan"},{"literal":"Kun Xu"},{"literal":"Yang Song"},{"literal":"Huawei Shen"},{"literal":"Xueqi Cheng"}],"issued":{"date-parts":[[2025,10,17]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2510.15522","URL":"https://arxiv.org/abs/2510.15522","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2510.15522/","first_public":"2025-10-17","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2510_24940","type":"article","title":"SemCoT: Accelerating Chain-of-Thought Reasoning through Semantically-Aligned Implicit Tokens","author":[{"literal":"Yinhan He"},{"literal":"Wendy Zheng"},{"literal":"Yaochen Zhu"},{"literal":"Zaiyi Zheng"},{"literal":"Lin Su"},{"literal":"Sriram Vasudevan"},{"literal":"Qi Guo"},{"literal":"Liangjie Hong"},{"literal":"Jundong Li"}],"issued":{"date-parts":[[2025,10,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2510.24940","URL":"https://arxiv.org/abs/2510.24940","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2510.24940/","first_public":"2025-10-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2510_25741","type":"article","title":"Scaling Latent Reasoning via Looped Language Models","author":[{"literal":"Rui-Jie Zhu"},{"literal":"Zixuan Wang"},{"literal":"Kai Hua"},{"literal":"Tianyu Zhang"},{"literal":"Ziniu Li"},{"literal":"Haoran Que"},{"literal":"Boyi Wei"},{"literal":"Zixin Wen"},{"literal":"Fan Yin"},{"literal":"He Xing"},{"literal":"Lu Li"},{"literal":"Jiajun Shi"},{"literal":"Kaijing Ma"},{"literal":"Shanda Li"},{"literal":"Taylor Kergan"},{"literal":"Andrew Smith"},{"literal":"Xingwei Qu"},{"literal":"Mude Hui"},{"literal":"Bohong Wu"},{"literal":"Qiyang Min"},{"literal":"Hongzhi Huang"},{"literal":"Xun Zhou"},{"literal":"Wei Ye"},{"literal":"Jiaheng Liu"},{"literal":"Jian Yang"},{"literal":"Yunfeng Shi"},{"literal":"Chenghua Lin"},{"literal":"Enduo Zhao"},{"literal":"Tianle Cai"},{"literal":"Ge Zhang"},{"literal":"Wenhao Huang"},{"literal":"Yoshua Bengio"},{"literal":"Jason Eshraghian"}],"issued":{"date-parts":[[2025,10,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2510.25741","URL":"https://arxiv.org/abs/2510.25741","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2510.25741/","first_public":"2025-10-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2511_07384","type":"article","title":"Teaching Pretrained Language Models to Think Deeper with Retrofitted Recurrence","author":[{"literal":"Sean McLeish"},{"literal":"Ang Li"},{"literal":"John Kirchenbauer"},{"literal":"Dayal Singh Kalra"},{"literal":"Brian R. Bartoldson"},{"literal":"Bhavya Kailkhura"},{"literal":"Avi Schwarzschild"},{"literal":"Jonas Geiping"},{"literal":"Tom Goldstein"},{"literal":"Micah Goldblum"}],"issued":{"date-parts":[[2025,11,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2511.07384","URL":"https://arxiv.org/abs/2511.07384","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2511.07384/","first_public":"2025-11-10","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2511_08577","type":"article","title":"Think-at-Hard: Dynamic Looped Transformers for Improved Reasoning","author":[{"literal":"Tianyu Fu"},{"literal":"Yichen You"},{"literal":"Zekai Chen"},{"literal":"Guohao Dai"},{"literal":"Huazhong Yang"},{"literal":"Yu Wang"}],"issued":{"date-parts":[[2025,11,11]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2511.08577","URL":"https://arxiv.org/abs/2511.08577","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2511.08577/","first_public":"2025-11-11","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2511_21581","type":"article","title":"Learning When to Stop: Adaptive Latent Reasoning via Reinforcement Learning","author":[{"literal":"Alex Ning"},{"literal":"Yen-Ling Kuo"},{"literal":"Gabe Gomes"}],"issued":{"date-parts":[[2025,11,26]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2511.21581","URL":"https://arxiv.org/abs/2511.21581","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2511.21581/","first_public":"2025-11-26","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2512_21711","type":"article","title":"Do Latent Tokens Think? A Causal and Adversarial Analysis of Chain-of-Continuous-Thought","author":[{"literal":"Yuyi Zhang"},{"literal":"Boyu Tang"},{"literal":"Tianjie Ju"},{"literal":"Sufeng Duan"},{"literal":"Gongshen Liu"}],"issued":{"date-parts":[[2025,12,25]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2512.21711","URL":"https://arxiv.org/abs/2512.21711","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2512.21711/","first_public":"2025-12-25","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2601_21358","type":"article","title":"Latent Chain-of-Thought as Planning: Decoupling Reasoning from Verbalization","author":[{"literal":"Jiecong Wang"},{"literal":"Hao Peng"},{"literal":"Zhanyi Wang"},{"literal":"Chunyang Liu"},{"literal":"Guanlin Wu"}],"issued":{"date-parts":[[2026,1,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2601.21358","URL":"https://arxiv.org/abs/2601.21358","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2601.21358/","first_public":"2026-01-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2601_21576","type":"article","title":"Chain Of Thought Compression: A Theoretical Analysis","author":[{"literal":"Juncai Li"},{"literal":"Ru Li"},{"literal":"Yuxiang Zhou"},{"literal":"Boxiang Ma"},{"literal":"Jeff Z. Pan"}],"issued":{"date-parts":[[2026,1,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2601.21576","URL":"https://arxiv.org/abs/2601.21576","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2601.21576/","first_public":"2026-01-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2601_21598","type":"article","title":"Beyond Imitation: Reinforcement Learning for Active Latent Planning","author":[{"literal":"Zhi Zheng"},{"literal":"Wee Sun Lee"}],"issued":{"date-parts":[[2026,1,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2601.21598","URL":"https://arxiv.org/abs/2601.21598","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2601.21598/","first_public":"2026-01-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2602_00449","type":"article","title":"Do Latent-CoT Models Think Step-by-Step? A Mechanistic Study on Sequential Reasoning Tasks","author":[{"literal":"Jia Liang"},{"literal":"Liangming Pan"}],"issued":{"date-parts":[[2026,1,31]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2602.00449","URL":"https://arxiv.org/abs/2602.00449","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2602.00449/","first_public":"2026-01-31","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2602_01148","type":"article","title":"Capabilities and Fundamental Limits of Latent Chain-of-Thought","author":[{"literal":"Jiaxuan Zou"},{"literal":"Yaozhong Xiong"},{"literal":"Yong Liu"}],"issued":{"date-parts":[[2026,2,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2602.01148","URL":"https://arxiv.org/abs/2602.01148","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2602.01148/","first_public":"2026-02-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2602_08332","type":"article","title":"Latent Reasoning with Supervised Thinking States","author":[{"literal":"Ido Amos"},{"literal":"Avi Caciularu"},{"literal":"Mor Geva"},{"literal":"Amir Globerson"},{"literal":"Jonathan Herzig"},{"literal":"Lior Shani"},{"literal":"Idan Szpektor"}],"issued":{"date-parts":[[2026,2,9]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2602.08332","URL":"https://arxiv.org/abs/2602.08332","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2602.08332/","first_public":"2026-02-09","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2602_11683","type":"article","title":"ThinkRouter: Efficient Reasoning via Routing Thinking between Latent and Discrete Spaces","author":[{"literal":"Xin Xu"},{"literal":"Tong Yu"},{"literal":"Xiang Chen"},{"literal":"Haoliang Wang"},{"literal":"Julian McAuley"},{"literal":"Saayan Mitra"}],"issued":{"date-parts":[[2026,2,12]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2602.11683","URL":"https://arxiv.org/abs/2602.11683","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2602.11683/","first_public":"2026-02-12","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2603_02023","type":"article","title":"PonderLM-3: Adaptive Token-Wise Pondering with Differentiable Masking","author":[{"literal":"He Li"},{"literal":"Feichen Song"},{"literal":"Boyi Zeng"},{"literal":"Shixiang Song"},{"literal":"Zhiqin John Xu"},{"literal":"Ziwei He"},{"literal":"Zhouhan Lin"}],"issued":{"date-parts":[[2026,3,2]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2603.02023","URL":"https://arxiv.org/abs/2603.02023","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2603.02023/","first_public":"2026-03-02","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2604_02371","type":"article","title":"Internalized Reasoning for Long-Context Visual Document Understanding","author":[{"literal":"Austin Veselka"}],"issued":{"date-parts":[[2026,3,31]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2604.02371","URL":"https://arxiv.org/abs/2604.02371","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2604.02371/","first_public":"2026-03-31","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2604_24881","type":"article","title":"Latent Agents: A Post-Training Procedure for Internalized Multi-Agent Debate","author":[{"literal":"John Seon Keun Yi"},{"literal":"Aaron Mueller"},{"literal":"Dokyun Lee"}],"issued":{"date-parts":[[2026,4,27]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2604.24881","URL":"https://arxiv.org/abs/2604.24881","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2604.24881/","first_public":"2026-04-27","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2605_28600","type":"article","title":"Transformers Provably Learn to Internalize Chain-of-Thought","author":[{"literal":"Yixiao Huang"},{"literal":"Hanlin Zhu"},{"literal":"Zixuan Wang"},{"literal":"Jiantao Jiao"},{"literal":"Stuart Russell"},{"literal":"Somayeh Sojoudi"},{"literal":"Song Mei"}],"issued":{"date-parts":[[2026,5,27]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2605.28600","URL":"https://arxiv.org/abs/2605.28600","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2605.28600/","first_public":"2026-05-27","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2606_01243","type":"article","title":"Unlocking the Black Box of Latent Reasoning: An Interpretability-Guided Approach to Intervention","author":[{"literal":"Shuochen Chang"},{"literal":"Tong Bai"},{"literal":"Xiaofeng Zhang"},{"literal":"Qianli Ma"},{"literal":"Qingyang Liu"},{"literal":"Zhaohe Liao"},{"literal":"Yibo Miao"},{"literal":"Li Niu"}],"issued":{"date-parts":[[2026,5,31]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2606.01243","URL":"https://arxiv.org/abs/2606.01243","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2606.01243/","first_public":"2026-05-31","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2606_12689","type":"article","title":"Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models","author":[{"literal":"Darpan Aswal"},{"literal":"Thomas Palmeira Ferraz"},{"literal":"Yongxin Zhou"},{"literal":"Maxime Peyrard"}],"issued":{"date-parts":[[2026,6,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2606.12689","URL":"https://arxiv.org/abs/2606.12689","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2606.12689/","first_public":"2026-06-10","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2606_13106","type":"article","title":"Demystifying Hidden-State Recurrence: Switchable Latent Reasoning with On-Policy Reinforcement Learning","author":[{"literal":"Jiayu Yang"},{"literal":"Chao Chen"},{"literal":"Shengen Wu"},{"literal":"Yinhong Liu"},{"literal":"Yuxuan Fan"},{"literal":"Lujundong Li"},{"literal":"Songning Lai"},{"literal":"Chengwei Qin"},{"literal":"Zhijiang Guo"}],"issued":{"date-parts":[[2026,6,11]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2606.13106","URL":"https://arxiv.org/abs/2606.13106","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2606.13106/","first_public":"2026-06-11","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2606_18206","type":"article","title":"Fixed-Point Reasoners: Stable and Adaptive Deep Looped Transformers","author":[{"literal":"Sajad Movahedi"},{"literal":"Vera Milovanović"},{"literal":"Shlomo Libo Feigin"},{"literal":"Alexander Theus"},{"literal":"Thomas Hofmann"},{"literal":"Valentina Boeva"},{"literal":"T. Konstantin Rusch"},{"literal":"Antonio Orvieto"}],"issued":{"date-parts":[[2026,6,16]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2606.18206","URL":"https://arxiv.org/abs/2606.18206","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2606.18206/","first_public":"2026-06-16","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2606_20075","type":"article","title":"What Makes Effective Supervision in Latent Chain-of-Thought: An Information-Theoretic Analysis","author":[{"literal":"Xinghao Chen"},{"literal":"Chak Tou Leong"},{"literal":"Wenjin Guo"},{"literal":"Jian Wang"},{"literal":"Wenjie Li"},{"literal":"Xiaoyu Shen"}],"issued":{"date-parts":[[2026,6,18]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2606.20075","URL":"https://arxiv.org/abs/2606.20075","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2606.20075/","first_public":"2026-06-18","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2607_00341","type":"article","title":"DiscoLoop: Looping Discrete Embeddings and Continuous Hidden States for Multi-hop Reasoning","author":[{"literal":"Hengyu Fu"},{"literal":"Tianyu Guo"},{"literal":"Zixuan Wang"},{"literal":"Hanlin Zhu"},{"literal":"Jason D. Lee"},{"literal":"Jiantao Jiao"},{"literal":"Stuart Russell"},{"literal":"Song Mei"}],"issued":{"date-parts":[[2026,7,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2607.00341","URL":"https://arxiv.org/abs/2607.00341","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2607.00341/","first_public":"2026-07-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2608_13570","type":"article","title":"Think in Latent, Explain in Language: Self-Explainable Latent Reasoning","author":[{"literal":"Dayuan Zhao"},{"literal":"Shengcao Cao"},{"literal":"Yu-Xiong Wang"},{"literal":"Liang-Yan Gui"}],"issued":{"date-parts":[[2026,7,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2608.13570","URL":"https://arxiv.org/abs/2608.13570","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2608.13570/","first_public":"2026-07-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2607_06648","type":"article","title":"Final Checkpoints Are Not Enough: Analyzing Latent Reasoning Faithfulness Along Training Trajectories","author":[{"literal":"Hengyu Jin"},{"literal":"Shu Yang"},{"literal":"Di Wang"}],"issued":{"date-parts":[[2026,7,7]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2607.06648","URL":"https://arxiv.org/abs/2607.06648","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2607.06648/","first_public":"2026-07-07","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2607_15178","type":"article","title":"T²MLR: Transformer with Temporal Middle-Layer Recurrence","author":[{"literal":"Ziyang Cai"},{"literal":"Xingyu Zhu"},{"literal":"Yihe Dong"},{"literal":"Yinghui He"},{"literal":"Sanjeev Arora"}],"issued":{"date-parts":[[2026,7,16]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2607.15178","URL":"https://arxiv.org/abs/2607.15178","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2607.15178/","first_public":"2026-07-16","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2607_16972","type":"article","title":"Training Continuous Chain of Thought Models: A Tale of Two Regimes","author":[{"literal":"Varun Yerram"},{"literal":"He He"},{"literal":"Eunsol Choi"}],"issued":{"date-parts":[[2026,7,18]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2607.16972","URL":"https://arxiv.org/abs/2607.16972","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2607.16972/","first_public":"2026-07-18","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2608_02585","type":"article","title":"GradCuit: Credit-Assigned Gradient Flow Enables Robust and Interpretable Test-Time Latent Reasoning","author":[{"literal":"Zhaoxin Yu"},{"literal":"Qi Shen"},{"literal":"Hengli Li"},{"literal":"Zhaowei Zhang"},{"literal":"Song-Chun Zhu"},{"literal":"Chi Zhang"},{"literal":"Zilong Zheng"}],"issued":{"date-parts":[[2026,8,3]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2608.02585","URL":"https://arxiv.org/abs/2608.02585","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2608.02585/","first_public":"2026-08-03","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2608_03624","type":"article","title":"LoopMTP: A looped transformer guided by latent multi-token prediction","author":[{"literal":"Behzad Shomali"},{"literal":"Markus Frey"},{"literal":"David Berghaus"},{"literal":"Joachim Koehler"},{"literal":"Mehdi Ali"}],"issued":{"date-parts":[[2026,8,4]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2608.03624","URL":"https://arxiv.org/abs/2608.03624","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2608.03624/","first_public":"2026-08-04","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2608_04928","type":"article","title":"Does Out-of-Sight Equal Out-of-Mind in CoT Monitorability?","author":[{"literal":"Pedro Ferreira"},{"literal":"Wilker Aziz"},{"literal":"Ivan Titov"}],"issued":{"date-parts":[[2026,8,5]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2608.04928","URL":"https://arxiv.org/abs/2608.04928","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2608.04928/","first_public":"2026-08-05","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2608_18222","type":"article","title":"Think Shallow, Solve Deep: Controlling Recurrent Dynamics for Reliable Test-Time Depth","author":[{"literal":"Ivan Viakhirev"},{"literal":"Kirill Borodin"},{"literal":"Amirah Almutairi"},{"literal":"Serguei Barannikov"},{"literal":"Maxim Abramov"},{"literal":"Grach Mkrtchian"}],"issued":{"date-parts":[[2026,8,18]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2608.18222","URL":"https://arxiv.org/abs/2608.18222","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2608.18222/","first_public":"2026-08-18","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2608_18230","type":"article","title":"Allocating Recurrent Compute in Looped Language Models","author":[{"literal":"Ruhai Lin"},{"literal":"Yiyang Guo"},{"literal":"Rui-Jie Zhu"},{"literal":"Hao Ye"},{"literal":"Jason K. Eshraghian"}],"issued":{"date-parts":[[2026,8,18]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2608.18230","URL":"https://arxiv.org/abs/2608.18230","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2608.18230/","first_public":"2026-08-18","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2608_29956","type":"article","title":"Detecting Hidden Chain-of-Thought in Large Language Models with Linguistic, Behavioral, and Mechanistic Indicators","author":[{"literal":"Armaan Singh"},{"literal":"Ryan Trinh Le"},{"literal":"Jasmine Kaur"},{"literal":"Abdullah Sultan"},{"literal":"Edward Lue Chee Lip"},{"literal":"Kiran Nijjer"},{"literal":"Adnan Ahmed"},{"literal":"Vasu Sharma"}],"issued":{"date-parts":[[2026,8,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2608.29956","URL":"https://arxiv.org/abs/2608.29956","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2608.29956/","first_public":"2026-08-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_22197","type":"article","title":"Dissecting Hierarchical Reasoning Models: A Mechanistic Study","author":[{"literal":"Leo Raphael Rodrigues"},{"literal":"Jian Kang"}],"issued":{"date-parts":[[2026,8,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.22197","URL":"https://arxiv.org/abs/2609.22197","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.22197/","first_public":"2026-08-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_00874","type":"article","title":"A Score Is Not a Policy: Measuring the Value of Adaptive Revision","author":[{"literal":"Ayushi Chadha"}],"issued":{"date-parts":[[2026,9,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.00874","URL":"https://arxiv.org/abs/2609.00874","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.00874/","first_public":"2026-09-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_01117","type":"article","title":"Latent Recurrent Thoughts: Recurrent Refinement of Proposed Latents for Reasoning with Frozen LLMs","author":[{"literal":"Zhaoliang Chen"},{"literal":"Jie Fu"}],"issued":{"date-parts":[[2026,9,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.01117","URL":"https://arxiv.org/abs/2609.01117","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.01117/","first_public":"2026-09-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_01343","type":"article","title":"SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers","author":[{"literal":"Shaowen Wang"},{"literal":"Ge Zhang"},{"literal":"Kairong Luo"},{"literal":"Yuhao Wu"},{"literal":"Shaofan Liu"},{"literal":"Jiaheng Liu"},{"literal":"Wenhao Huang"},{"literal":"Shen Yan"},{"literal":"Jian Li"}],"issued":{"date-parts":[[2026,9,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.01343","URL":"https://arxiv.org/abs/2609.01343","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.01343/","first_public":"2026-09-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_01449","type":"article","title":"Diffusion as a Training Curriculum for Timestep-Free Iterative Reasoning","author":[{"literal":"Mariia Drozdova"},{"literal":"Aidan Sirbu"},{"literal":"Pietro Miotti"},{"literal":"Robert Obryk"},{"literal":"Mayalen Etcheverry"},{"literal":"Eyvind Niklasson"},{"literal":"Blake Richards"}],"issued":{"date-parts":[[2026,9,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.01449","URL":"https://arxiv.org/abs/2609.01449","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.01449/","first_public":"2026-09-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_01924","type":"article","title":"Looped Transformers under the Jacobian Lens: Does the Global Workspace Survive Recurrence?","author":[{"literal":"Wenlong Wang"},{"literal":"Fergal Reid"}],"issued":{"date-parts":[[2026,9,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.01924","URL":"https://arxiv.org/abs/2609.01924","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.01924/","first_public":"2026-09-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_03379","type":"article","title":"RecurTrace: Adaptive Latent Reasoning with Loop-Time Memory","author":[{"literal":"Yuxiang Wang"},{"literal":"Kunyu Feng"},{"literal":"Yingda Shen"},{"literal":"Haoning Xu"},{"literal":"Junyu Wang"},{"literal":"Zhizheng Wu"}],"issued":{"date-parts":[[2026,9,3]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.03379","URL":"https://arxiv.org/abs/2609.03379","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.03379/","first_public":"2026-09-03","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_03807","type":"article","title":"Almost Free State Prediction Separation","author":[{"literal":"John Langford"},{"literal":"Nathan Godey"},{"literal":"Giovanni Monea"},{"literal":"Yoav Artzi"},{"literal":"Harry Dong"},{"literal":"Ying Fan"},{"literal":"Gustavo de Rosa"},{"literal":"Zheng Zhan"}],"issued":{"date-parts":[[2026,9,3]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.03807","URL":"https://arxiv.org/abs/2609.03807","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.03807/","first_public":"2026-09-03","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_04489","type":"article","title":"Towards Understanding Pause Token Fine-Tuning Dynamics: A Mode Retention Perspective","author":[{"literal":"Jaehyeon Kim"},{"literal":"Suhwan Kim"},{"literal":"Nakyung Lee"},{"literal":"Yeongoon Kim"},{"literal":"Jimin Seo"},{"literal":"Giho Lee"},{"literal":"Jungwoo Lee"}],"issued":{"date-parts":[[2026,9,3]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.04489","URL":"https://arxiv.org/abs/2609.04489","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.04489/","first_public":"2026-09-03","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_04963","type":"article","title":"Fractal basins trap latent reasoning","author":[{"literal":"Jeffrey Lai"},{"literal":"Anthony Bao"},{"literal":"John Quinn"},{"literal":"William Gilpin"}],"issued":{"date-parts":[[2026,9,4]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.04963","URL":"https://arxiv.org/abs/2609.04963","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.04963/","first_public":"2026-09-04","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_05988","type":"article","title":"DART: Distributional Adversarial Recurrent Training for Algorithm Learning","author":[{"literal":"Hieu Tran Bao"},{"literal":"Phung Thanh Dang"},{"literal":"Pham Quang Nhat Minh"},{"literal":"Hoang Thanh Tung"}],"issued":{"date-parts":[[2026,9,5]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.05988","URL":"https://arxiv.org/abs/2609.05988","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.05988/","first_public":"2026-09-05","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_09928","type":"article","title":"Structural Process Supervision for Latent Chain-of-Thought Reasoning","author":[{"literal":"Yiqi Li"},{"literal":"Xu Chen"},{"literal":"Chen Ju"},{"literal":"Jiangchao Yao"},{"literal":"Zhaoyang Li"},{"literal":"Jinsong Lan"},{"literal":"Xiaoyong Zhu"},{"literal":"Yu Wang"}],"issued":{"date-parts":[[2026,9,9]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.09928","URL":"https://arxiv.org/abs/2609.09928","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.09928/","first_public":"2026-09-09","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_13060","type":"article","title":"CanvasAnneal: Curriculum Reinforcement Learning for Diffusion Language Models","author":[{"literal":"Blake Olson"},{"literal":"Yuhang Song"},{"literal":"Emmett McQuinn"},{"literal":"Yuan Shangguan"}],"issued":{"date-parts":[[2026,9,11]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.13060","URL":"https://arxiv.org/abs/2609.13060","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.13060/","first_public":"2026-09-11","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_15160","type":"article","title":"T-LoopFormer: Token-Level Elastic-Depth Looped Transformers with Dynamic Routing","author":[{"literal":"Mingqian Yu"},{"literal":"Wenpeng Zhang"},{"literal":"Shaobo Cui"},{"literal":"Peilin Zhao"}],"issued":{"date-parts":[[2026,9,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.15160","URL":"https://arxiv.org/abs/2609.15160","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.15160/","first_public":"2026-09-14","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_16372","type":"article","title":"Register Tokens for Bounded-State Reasoning in Diffusion Language Models","author":[{"literal":"Albert Ge"},{"literal":"Chandan Singh"},{"literal":"Yufan Zhuang"},{"literal":"Xiaodong Liu"},{"literal":"Jianfeng Gao"},{"literal":"Frederic Sala"}],"issued":{"date-parts":[[2026,9,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.16372","URL":"https://arxiv.org/abs/2609.16372","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.16372/","first_public":"2026-09-14","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_16665","type":"article","title":"Right Direction, Wrong Step: Geometric Analysis of Finite-Step Failure in Looped Transformers","author":[{"literal":"Zhihao Guo"},{"literal":"Zonghan Wu"},{"literal":"Haizhou Du"},{"literal":"Huan Huo"},{"literal":"Yilei Shao"},{"literal":"Athanasios V. Vasilakos"},{"literal":"Qingsong Wen"}],"issued":{"date-parts":[[2026,9,15]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.16665","URL":"https://arxiv.org/abs/2609.16665","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.16665/","first_public":"2026-09-15","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_17184","type":"article","title":"LoopSpec: Pipelined Self-Speculative Decoding for Looped Transformers","author":[{"literal":"SangLyul Cho"},{"literal":"Langqing Cui"},{"literal":"Sehoon Kim"},{"literal":"Dongsu Han"},{"literal":"Insu Han"}],"issued":{"date-parts":[[2026,9,15]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.17184","URL":"https://arxiv.org/abs/2609.17184","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.17184/","first_public":"2026-09-15","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_19107","type":"article","title":"How Model Growth, Recursion, and Boundary Operators Influence Scaling Exponents","author":[{"literal":"Zixi Chen"},{"literal":"Akshay Vegesna"},{"literal":"Samip Dahal"},{"literal":"Andrew Gordon Wilson"}],"issued":{"date-parts":[[2026,9,16]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.19107","URL":"https://arxiv.org/abs/2609.19107","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.19107/","first_public":"2026-09-16","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_19521","type":"article","title":"Storing Is Not Remembering: LSTM-UT and Bounded Gated Memory for Looped Transformers","author":[{"literal":"Aras Kavuncu"},{"literal":"Muhammad Burhan Hafez"}],"issued":{"date-parts":[[2026,9,17]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.19521","URL":"https://arxiv.org/abs/2609.19521","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.19521/","first_public":"2026-09-17","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_19717","type":"article","title":"Learn Your Own Thoughts: Abstract Token Curriculum","author":[{"literal":"Khashayar Gatmiry"},{"literal":"Avrajit Ghosh"},{"literal":"Parsa Mirtaheri"},{"literal":"Jason D. Lee"},{"literal":"Nika Haghtalab"},{"literal":"Emmanuel Abbe"},{"literal":"Peter Bartlett"}],"issued":{"date-parts":[[2026,9,17]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.19717","URL":"https://arxiv.org/abs/2609.19717","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.19717/","first_public":"2026-09-17","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_19878","type":"article","title":"Uni-LaDiR: Latent Diffusion Unifies Multimodal Reasoning","author":[{"literal":"Haoqiang Kang"},{"literal":"Yizhe Zhang"},{"literal":"Nikki Lijing Kuang"},{"literal":"Jiatao Gu"},{"literal":"Yian Ma"},{"literal":"Lianhui Qin"}],"issued":{"date-parts":[[2026,9,17]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.19878","URL":"https://arxiv.org/abs/2609.19878","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.19878/","first_public":"2026-09-17","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_20398","type":"article","title":"Schema-Anchored Latent Reasoning for Semantic Parsing-Based Knowledge Base Question Answering","author":[{"literal":"Guangze Gao"},{"literal":"Zixuan Li"},{"literal":"Sikui Zhang"},{"literal":"Chunfeng Yuan"},{"literal":"Wenjuan Li"},{"literal":"Bing Li"},{"literal":"Xiaolong Jin"},{"literal":"Weiming Hu"}],"issued":{"date-parts":[[2026,9,17]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.20398","URL":"https://arxiv.org/abs/2609.20398","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.20398/","first_public":"2026-09-17","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_20981","type":"article","title":"CaLR: Causal Latent Revision for Robust Diffusion Reasoning","author":[{"literal":"Wei Cai"},{"literal":"Jian Zhao"},{"literal":"Yuchen Yuan"},{"literal":"Xuelong Li"}],"issued":{"date-parts":[[2026,9,17]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.20981","URL":"https://arxiv.org/abs/2609.20981","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.20981/","first_public":"2026-09-17","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_21605","type":"article","title":"Trading Depth for Time in Recurrent Transformers","author":[{"literal":"Zeyi Huang"},{"literal":"Xuehai He"},{"literal":"Yelong Shen"},{"literal":"Yong Jae Lee"}],"issued":{"date-parts":[[2026,9,18]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.21605","URL":"https://arxiv.org/abs/2609.21605","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.21605/","first_public":"2026-09-18","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_21662","type":"article","title":"When Steering Fails in Latent Reasoning: A Latent-to-Language Transition Gap","author":[{"literal":"Gaoxiang Huang"},{"literal":"Lei Qi"}],"issued":{"date-parts":[[2026,9,18]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.21662","URL":"https://arxiv.org/abs/2609.21662","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.21662/","first_public":"2026-09-18","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_23033","type":"article","title":"WaveFront Decoding: Parallelized Self-Speculative Decoding for Looped Language Models","author":[{"literal":"Hyeongju Ha"},{"literal":"Jae-Joon Kim"}],"issued":{"date-parts":[[2026,9,19]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.23033","URL":"https://arxiv.org/abs/2609.23033","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.23033/","first_public":"2026-09-19","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_24196","type":"article","title":"LoopCD: Loop-wise Contrastive Decoding for Improving Reasoning in Looped Language Models","author":[{"literal":"Byeongho Yu"},{"literal":"Junhyuk So"},{"literal":"Eunhyeok Park"}],"issued":{"date-parts":[[2026,9,21]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.24196","URL":"https://arxiv.org/abs/2609.24196","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.24196/","first_public":"2026-09-21","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_24346","type":"article","title":"LADDER: Graph-Guided Diffusion Language Models for Efficient Multi-Hop Reasoning","author":[{"literal":"Senlei Zhang"},{"literal":"Linhao Luo"},{"literal":"Qian-Wen Zhang"},{"literal":"Siyu An"},{"literal":"Junnan Dong"},{"literal":"Shuhao Zhang"},{"literal":"Xing Sun"}],"issued":{"date-parts":[[2026,9,21]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.24346","URL":"https://arxiv.org/abs/2609.24346","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.24346/","first_public":"2026-09-21","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_27220","type":"article","title":"LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Models","author":[{"literal":"Guoshenghui Zhao"},{"literal":"Tan Yu"},{"literal":"Weijie Zhao"}],"issued":{"date-parts":[[2026,9,23]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.27220","URL":"https://arxiv.org/abs/2609.27220","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.27220/","first_public":"2026-09-23","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_27373","type":"article","title":"Attention Routing Stabilizes Early: Working-Set Inference for Recurrent Language Models","author":[{"literal":"Ke Wan"},{"literal":"Chen Chen"}],"issued":{"date-parts":[[2026,9,23]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.27373","URL":"https://arxiv.org/abs/2609.27373","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.27373/","first_public":"2026-09-23","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_28272","type":"article","title":"Towards Efficient Reasoning: Learning Causal Shortcuts for Diffusion Language Models","author":[{"literal":"Dian Jin"},{"literal":"Kairong Han"},{"literal":"Baohong Li"},{"literal":"Xinpeng Dong"},{"literal":"Zijing Hu"},{"literal":"Nuanqiao Shan"},{"literal":"Fei Wu"},{"literal":"Kun Kuang"}],"issued":{"date-parts":[[2026,9,23]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.28272","URL":"https://arxiv.org/abs/2609.28272","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.28272/","first_public":"2026-09-23","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_29102","type":"article","title":"ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks","author":[{"literal":"Zeyu Michael Li"},{"literal":"William Xingxu Chen"},{"literal":"Bingshuo Qian"},{"literal":"Jiayin Liu"},{"literal":"Xiang Cheng"}],"issued":{"date-parts":[[2026,9,24]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.29102","URL":"https://arxiv.org/abs/2609.29102","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.29102/","first_public":"2026-09-24","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_08820","type":"article","title":"Task-Oriented Key-Layer KV Communication for Efficient Latent Multi-Agent Collaboration","author":[{"literal":"Dongsen Zhang"},{"literal":"Peipei Li"},{"literal":"Zekun Li"},{"literal":"Wenjun Xu"}],"issued":{"date-parts":[[2026,9,24]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.08820","URL":"https://arxiv.org/abs/2610.08820","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.08820/","first_public":"2026-09-24","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_30783","type":"article","title":"Skip the Talk, Re-Focus on Vision: Latent Reasoning for Reasoning Segmentation in Multimodal Large Language Models","author":[{"literal":"Tianhang Guo"},{"literal":"Yulin He"},{"literal":"Wei Chen"},{"literal":"Wenjuan Zhou"},{"literal":"Yuhang Li"},{"literal":"Xinbiao Gan"}],"issued":{"date-parts":[[2026,9,25]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.30783","URL":"https://arxiv.org/abs/2609.30783","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.30783/","first_public":"2026-09-25","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_33134","type":"article","title":"Ceiling of a Task: When Can a Transformer Succeed Without Its Chain of Thought?","author":[{"literal":"Jiashu He"},{"literal":"Jinxuan Fan"},{"literal":"Xiao Xiao"},{"literal":"Radu Marculescu"},{"literal":"Alejandro Ribeiro"}],"issued":{"date-parts":[[2026,9,27]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.33134","URL":"https://arxiv.org/abs/2609.33134","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.33134/","first_public":"2026-09-27","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_33144","type":"article","title":"Beyond the Training Horizon: Mechanisms and Limits of Length Generalization in Looped Transformers","author":[{"literal":"Jia Liang"},{"literal":"Xi Jin"},{"literal":"Liangming Pan"}],"issued":{"date-parts":[[2026,9,27]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.33144","URL":"https://arxiv.org/abs/2609.33144","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.33144/","first_public":"2026-09-27","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_33270","type":"article","title":"Structured Sparse Memory for Recurrent Reasoning","author":[{"literal":"Zixuan Zhao"},{"literal":"Samuel Wheeler"},{"literal":"Neil Getty"},{"literal":"Xiaotian Duan"},{"literal":"Rick Stevens"},{"literal":"Fangfang Xia"}],"issued":{"date-parts":[[2026,9,27]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.33270","URL":"https://arxiv.org/abs/2609.33270","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.33270/","first_public":"2026-09-27","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_33271","type":"article","title":"Next Thoughts Are Distributions: Generative Autoregressive Reasoning in the Latent Space","author":[{"literal":"Yang Li"},{"literal":"Yi Wang"},{"literal":"Shiyuan Huang"},{"literal":"Yang Liu"},{"literal":"Hao Wang"},{"literal":"Chengzhi Mao"}],"issued":{"date-parts":[[2026,9,27]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.33271","URL":"https://arxiv.org/abs/2609.33271","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.33271/","first_public":"2026-09-27","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_33402","type":"article","title":"VaME: Exploring Variational Latent Reasoning for Multimodal Embeddings","author":[{"literal":"Peixi Wu"},{"literal":"Mingzhou Jiang"},{"literal":"Feipeng Ma"},{"literal":"Biao Yang"},{"literal":"Yunhao Zhou"},{"literal":"Wei Yuan"},{"literal":"Bosong Chai"},{"literal":"Huizu Lin"},{"literal":"Jie Chen"},{"literal":"Zhangchi Hu"},{"literal":"Fan Yang"},{"literal":"Wenwu Ou"},{"literal":"Hebei Li"},{"literal":"Xiaoyan Sun"}],"issued":{"date-parts":[[2026,9,27]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.33402","URL":"https://arxiv.org/abs/2609.33402","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.33402/","first_public":"2026-09-27","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_34218","type":"article","title":"Loop Dropout: Regularizing Shared Updates in Looped Language Models","author":[{"literal":"Zirui Zhu"},{"literal":"Hailun Xu"},{"literal":"Xuanlei Zhao"},{"literal":"Yong Liu"},{"literal":"Yingxuan Ren"},{"literal":"Kanchan Sarkar"},{"literal":"Kun Xu"},{"literal":"Yang You"}],"issued":{"date-parts":[[2026,9,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.34218","URL":"https://arxiv.org/abs/2609.34218","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.34218/","first_public":"2026-09-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_34408","type":"article","title":"Distilling Visual Reasoning into Text Space","author":[{"literal":"Wenhan Yang"},{"literal":"Nilay Naharas"},{"literal":"Ali Payani"},{"literal":"Baharan Mirzasoleiman"}],"issued":{"date-parts":[[2026,9,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.34408","URL":"https://arxiv.org/abs/2609.34408","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.34408/","first_public":"2026-09-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_34488","type":"article","title":"FlexLoop: Depth-Elastic Looped Policies for Adaptive Test-Time Computation in Deep RL","author":[{"literal":"Xun Wang"},{"literal":"Ruishuo Chen"},{"literal":"Yu Chen"},{"literal":"Zhuoran Li"},{"literal":"Longbo Huang"}],"issued":{"date-parts":[[2026,9,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.34488","URL":"https://arxiv.org/abs/2609.34488","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.34488/","first_public":"2026-09-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_34538","type":"article","title":"Shallow Queries, Mature Values: Depth-Asynchronous Self-Speculation for Looped Transformers","author":[{"literal":"Guanghao Li"},{"literal":"Zihan Su"},{"literal":"Hao Yu"},{"literal":"Jinyang Jiang"},{"literal":"Tao Ren"},{"literal":"Zehao Li"},{"literal":"Feng Lu"},{"literal":"Ming Tang"},{"literal":"Chun Yuan"}],"issued":{"date-parts":[[2026,9,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.34538","URL":"https://arxiv.org/abs/2609.34538","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.34538/","first_public":"2026-09-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_35472","type":"article","title":"Why Deterministic PRM Guidance Underperforms in Discrete Diffusion Reasoning","author":[{"literal":"Yan Zhan"},{"literal":"Shaobo Liu"},{"literal":"Zhijun Gao"}],"issued":{"date-parts":[[2026,9,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.35472","URL":"https://arxiv.org/abs/2609.35472","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.35472/","first_public":"2026-09-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_35694","type":"article","title":"Reasoning with Continuous Latent Diffusion","author":[{"literal":"Xiang Cheng"}],"issued":{"date-parts":[[2026,9,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.35694","URL":"https://arxiv.org/abs/2609.35694","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.35694/","first_public":"2026-09-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_35748","type":"article","title":"Improving Test-Time Scaling with Adaptive Looped Transformers","author":[{"literal":"Yichen You"},{"literal":"Tianyu Fu"},{"literal":"Aosong Feng"},{"literal":"Xingtai Lv"},{"literal":"Xuefei Ning"},{"literal":"Ning Ding"},{"literal":"Yu Wang"}],"issued":{"date-parts":[[2026,9,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.35748","URL":"https://arxiv.org/abs/2609.35748","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.35748/","first_public":"2026-09-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_35751","type":"article","title":"How to Loop MoE: Flatten the Experts, Untie the Attention","author":[{"literal":"Shouren Wang"},{"literal":"Chuang Ma"},{"literal":"Mohsen Hariri"},{"literal":"Debargha Ganguly"},{"literal":"Wang Yang"},{"literal":"Xiaoqing Tong"},{"literal":"Qianying Liu"},{"literal":"Xiaotian Han"},{"literal":"Vipin Chaudhary"}],"issued":{"date-parts":[[2026,9,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.35751","URL":"https://arxiv.org/abs/2609.35751","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.35751/","first_public":"2026-09-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_08833","type":"article","title":"CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models","author":[{"literal":"Yue Wu"},{"literal":"Qinghe Zhang"},{"literal":"Yu Zhang"},{"literal":"Jian Huang"}],"issued":{"date-parts":[[2026,9,28]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.08833","URL":"https://arxiv.org/abs/2610.08833","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.08833/","first_public":"2026-09-28","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_36452","type":"article","title":"Reliable Parallel Decoding in Masked Diffusion Language Models","author":[{"literal":"Zhenghao He"},{"literal":"Bohan Liu"},{"literal":"Guangzhi Xiong"},{"literal":"Aidong Zhang"}],"issued":{"date-parts":[[2026,9,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.36452","URL":"https://arxiv.org/abs/2609.36452","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.36452/","first_public":"2026-09-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_36636","type":"article","title":"What Makes Recurrence Effective in Looped Language Models?","author":[{"literal":"Xinlin Zhuang"},{"literal":"Siyuan Wang"},{"literal":"Imran Razzak"},{"literal":"Weiyang Liu"}],"issued":{"date-parts":[[2026,9,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.36636","URL":"https://arxiv.org/abs/2609.36636","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.36636/","first_public":"2026-09-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_36653","type":"article","title":"Scheduling Recursive Reasoning in Looped Transformers","author":[{"literal":"Boyuan Wang"},{"literal":"Chengyao Yu"},{"literal":"Jiaxi Ren"},{"literal":"Hongxin Wei"},{"literal":"Bingyi Jing"},{"literal":"Yuxin Tao"}],"issued":{"date-parts":[[2026,9,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.36653","URL":"https://arxiv.org/abs/2609.36653","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.36653/","first_public":"2026-09-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_37351","type":"article","title":"Port-Hamiltonian Latent Deliberation: Mitigating the Deliberation Drift Cliff in Test-Time Compute Scaling","author":[{"literal":"Zeyu Jia"}],"issued":{"date-parts":[[2026,9,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.37351","URL":"https://arxiv.org/abs/2609.37351","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.37351/","first_public":"2026-09-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_37379","type":"article","title":"Looped Transformers as Optimizers","author":[{"literal":"Yulong Huang"},{"literal":"Chen Jiang"},{"literal":"Zhanpeng Zhou"},{"literal":"Hongtao Zhang"},{"literal":"Tianyu Li"},{"literal":"Tianyu He"},{"literal":"Xiangyu Zhang"},{"literal":"Bojun Cheng"}],"issued":{"date-parts":[[2026,9,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.37379","URL":"https://arxiv.org/abs/2609.37379","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.37379/","first_public":"2026-09-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_37432","type":"article","title":"Looped Actor: Depth-Recurrent Reasoning Models for Reinforcement Learning","author":[{"literal":"T. Konstantin Rusch"},{"literal":"Tim Seyde"},{"literal":"Jared Boyer"},{"literal":"Zach J. Patterson"},{"literal":"Daniela Rus"}],"issued":{"date-parts":[[2026,9,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.37432","URL":"https://arxiv.org/abs/2609.37432","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.37432/","first_public":"2026-09-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_38426","type":"article","title":"LoopVL: Recurrent Visual Intelligence","author":[{"literal":"Zhe Qian"},{"literal":"Ziyang Gong"},{"literal":"Zhongxing Xu"},{"literal":"Hehan Li"},{"literal":"Zhonghua Wang"},{"literal":"Hanmeng Liu"},{"literal":"Liwei Qian"},{"literal":"Xin Pei"},{"literal":"Peizhi Xu"},{"literal":"Fei Luo"},{"literal":"Mingxuan Wang"},{"literal":"Xue Yang"},{"literal":"Shiwei Liu"},{"literal":"Yanbiao Ma"},{"literal":"Junchi Yan"},{"literal":"Jungong Han"}],"issued":{"date-parts":[[2026,9,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.38426","URL":"https://arxiv.org/abs/2609.38426","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.38426/","first_public":"2026-09-29","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_38806","type":"article","title":"Blackboard Intelligence Can Surpass Autoregressive on Globally Constrained Problems","author":[{"literal":"Woosang Jeon"},{"literal":"Jaeyeon Kim"},{"literal":"Sham Kakade"},{"literal":"Yilun Du"},{"literal":"Amrit Singh Bedi"},{"literal":"Arun Kumar Chithanar"},{"literal":"Chul Lee"},{"literal":"Taehyeong Kim"},{"literal":"Sitan Chen"}],"issued":{"date-parts":[[2026,9,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.38806","URL":"https://arxiv.org/abs/2609.38806","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.38806/","first_public":"2026-09-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_39740","type":"article","title":"LatentHarness: Learning Latent Actions for Memory and Reasoning via Counterfactual Policy Distillation","author":[{"literal":"Xiaoqiang Wang"},{"literal":"Suyuchen Wang"},{"literal":"Bang Liu"}],"issued":{"date-parts":[[2026,9,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.39740","URL":"https://arxiv.org/abs/2609.39740","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.39740/","first_public":"2026-09-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_39892","type":"article","title":"Shared Weights, Selected Computations: How Looped Transformers Route What Each Loop Does","author":[{"literal":"Jiaju Wu"},{"literal":"Yi Hu"},{"literal":"Muhan Zhang"}],"issued":{"date-parts":[[2026,9,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.39892","URL":"https://arxiv.org/abs/2609.39892","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.39892/","first_public":"2026-09-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_39967","type":"article","title":"What Limits Recursive Reasoning Models: Optimization, Architecture and Test-Time Scaling","author":[{"literal":"Yuliana Shakhvalieva"},{"literal":"Dmitrii Kharchev"},{"literal":"Viacheslav Bezrukov"},{"literal":"Inessa Fedorova"},{"literal":"Dmitry Bocharov"},{"literal":"Ivan Oseledets"},{"literal":"Valerii Ternovskii"}],"issued":{"date-parts":[[2026,9,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.39967","URL":"https://arxiv.org/abs/2609.39967","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.39967/","first_public":"2026-09-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2609_40316","type":"article","title":"Scaling Laws for Looped Mixture of Experts","author":[{"literal":"Yanbei Chen"},{"literal":"Anirudh Goyal"},{"literal":"Raghuraman Krishnamoorthi"}],"issued":{"date-parts":[[2026,9,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2609.40316","URL":"https://arxiv.org/abs/2609.40316","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2609.40316/","first_public":"2026-09-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_00661","type":"article","title":"Exploring More, Reasoning Better: Stepwise Risk-Sensitive GRPO for Diffusion Language Models","author":[{"literal":"Yue Yu"},{"literal":"Bowen Zuo"},{"literal":"David Crandall"},{"literal":"Yinglun Zhu"},{"literal":"Dongruo Zhou"}],"issued":{"date-parts":[[2026,9,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.00661","URL":"https://arxiv.org/abs/2610.00661","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.00661/","first_public":"2026-09-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_00673","type":"article","title":"Closing the Loop: Practical Training Recipes for Looped Language Models","author":[{"literal":"Andrei Marchenko"},{"literal":"Viacheslav Bezrukov"},{"literal":"Oleg Kashurin"},{"literal":"Inessa Fedorova"},{"literal":"Dmitry Bocharov"},{"literal":"Yuliana Shakhvalieva"},{"literal":"Maria Tikhonova"},{"literal":"Valerii Ternovskii"}],"issued":{"date-parts":[[2026,9,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.00673","URL":"https://arxiv.org/abs/2610.00673","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.00673/","first_public":"2026-09-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_14030","type":"article","title":"Rotate the Loop: Blockwise Conformality for Looped Transformers","author":[{"literal":"Alexey Zemtsov"},{"literal":"Albina Klepach"},{"literal":"Vadim Lomshakov"},{"literal":"Andrei Polubarov"},{"literal":"Nikita Lyubaykin"},{"literal":"Alexander Derevyagin"},{"literal":"Vladislav Kurenkov"},{"literal":"Alexander Nikulin"}],"issued":{"date-parts":[[2026,9,30]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.14030","URL":"https://arxiv.org/abs/2610.14030","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.14030/","first_public":"2026-09-30","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_01153","type":"article","title":"Looping Beyond Twice: A Scalable Recipe for Looped Mixture-of-Experts","author":[{"literal":"Di He"},{"literal":"Pengxiang Li"},{"literal":"Da Chang"},{"literal":"Qingyan Meng"},{"literal":"Lu Yin"},{"literal":"Shiwei Liu"}],"issued":{"date-parts":[[2026,10,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.01153","URL":"https://arxiv.org/abs/2610.01153","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.01153/","first_public":"2026-10-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_01177","type":"article","title":"Temporally-Resolved Token Attribution Reveals the Generation Dynamics of Diffusion Language Models","author":[{"literal":"Darpan Aswal"},{"literal":"Céline Hudelot"}],"issued":{"date-parts":[[2026,10,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.01177","URL":"https://arxiv.org/abs/2610.01177","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.01177/","first_public":"2026-10-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_02185","type":"article","title":"Decoding Looped Transformers Better for (Almost) Free","author":[{"literal":"Weihao Liu"},{"literal":"Huangjie Zheng"},{"literal":"Tianrong Chen"},{"literal":"Rohit Dilip"},{"literal":"Richard He Bai"},{"literal":"Yizhu Jiao"},{"literal":"Yuyang Wang"},{"literal":"Ruixiang Zhang"}],"issued":{"date-parts":[[2026,10,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.02185","URL":"https://arxiv.org/abs/2610.02185","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.02185/","first_public":"2026-10-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_02193","type":"article","title":"Hierarchical Continuous Diffusion Language Models","author":[{"literal":"Hui Ren"},{"literal":"Zihan Li"},{"literal":"Chang Liu"},{"literal":"Huidong Liu"},{"literal":"Alexander Schwing"}],"issued":{"date-parts":[[2026,10,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.02193","URL":"https://arxiv.org/abs/2610.02193","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.02193/","first_public":"2026-10-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_02383","type":"article","title":"The Surprising Effectiveness of Shared Memory in Looped Transformers","author":[{"literal":"Giovanni Monea"},{"literal":"Keshav Ramji"},{"literal":"Yousef El-Kurdi"},{"literal":"Luis A. Lastras"},{"literal":"Yoav Artzi"},{"literal":"Nathan Godey"},{"literal":"Ramón Fernandez Astudillo"}],"issued":{"date-parts":[[2026,10,1]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.02383","URL":"https://arxiv.org/abs/2610.02383","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.02383/","first_public":"2026-10-01","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_02665","type":"article","title":"Large Language Continuous Diffusion Models","author":[{"literal":"Zhihan Yang"},{"literal":"Wei Guo"},{"literal":"Jean-Marie Lemercier"},{"literal":"Simon Welker"},{"literal":"Yonggan Fu"},{"literal":"Mohammad Mahdi Kamani"},{"literal":"Sajad Norouzi"},{"literal":"Julius Berner"},{"literal":"Tomas Geffner"},{"literal":"Karsten Kreis"},{"literal":"Yongxin Chen"},{"literal":"Molei Tao"},{"literal":"John Thickstun"},{"literal":"Pavlo Molchanov"},{"literal":"Ante Jukić"},{"literal":"Arash Vahdat"},{"literal":"Morteza Mardani"}],"issued":{"date-parts":[[2026,10,2]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.02665","URL":"https://arxiv.org/abs/2610.02665","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.02665/","first_public":"2026-10-02","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_02741","type":"article","title":"On the Chain-of-Thought Monitorability of Looped Language Models","author":[{"literal":"Han Wang"},{"literal":"Ishwar B Balappanawar"},{"literal":"Huan Zhang"}],"issued":{"date-parts":[[2026,10,2]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.02741","URL":"https://arxiv.org/abs/2610.02741","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.02741/","first_public":"2026-10-02","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_03665","type":"article","title":"Pivot-SD: Efficient Self-Distillation for Masked Diffusion Language Models","author":[{"literal":"Seo Hyun Kim"},{"literal":"Sunwoo Hong"},{"literal":"Younwoo Choi"},{"literal":"Chen-Hao Chao"},{"literal":"Se-Young Yun"},{"literal":"Rahul G. Krishnan"}],"issued":{"date-parts":[[2026,10,2]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.03665","URL":"https://arxiv.org/abs/2610.03665","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.03665/","first_public":"2026-10-02","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_06949","type":"article","title":"AdaLoop: Adaptive-Depth Latent Reasoning for Audio Language Models","author":[{"literal":"Lee Seung-woo"},{"literal":"Bowen Qi"}],"issued":{"date-parts":[[2026,10,3]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.06949","URL":"https://arxiv.org/abs/2610.06949","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.06949/","first_public":"2026-10-03","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_04864","type":"article","title":"Enhancing Long-Video VLM Embeddings with Query-Aware Streaming Latent Reasoning","author":[{"literal":"Haozhe Chi"},{"literal":"Song Jin"},{"literal":"Yang Jin"},{"literal":"Yadong Mu"}],"issued":{"date-parts":[[2026,10,4]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.04864","URL":"https://arxiv.org/abs/2610.04864","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.04864/","first_public":"2026-10-04","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_06666","type":"article","title":"What Matters for Latent Reasoning with Flow Matching","author":[{"literal":"Yassine Ouali"},{"literal":"Adrian Bulat"},{"literal":"Georgios Tzimiropoulos"}],"issued":{"date-parts":[[2026,10,5]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.06666","URL":"https://arxiv.org/abs/2610.06666","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.06666/","first_public":"2026-10-05","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_06833","type":"article","title":"Towards Looped Models Done Right, Part II: Rethinking at Fixed Points","author":[{"literal":"Benhao Huang"},{"literal":"Chufan Shi"},{"literal":"Junlin Chen"},{"literal":"Shicheng Wen"},{"literal":"Zhengzhong Liu"},{"literal":"Eric Xing"},{"literal":"Xuezhe Ma"}],"issued":{"date-parts":[[2026,10,5]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.06833","URL":"https://arxiv.org/abs/2610.06833","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.06833/","first_public":"2026-10-05","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_07175","type":"article","title":"CALR: Continuous Anchored Latent Reasoning via Render-of-Thought Compression","author":[{"literal":"Zhaoyang Wei"},{"literal":"Bowen Jiang"},{"literal":"Yanchao Hao"},{"literal":"Wenchao Ding"},{"literal":"Zheng Wei"},{"literal":"Shaocheng Wu"},{"literal":"Zhenjun Han"},{"literal":"Jianbin Jiao"}],"issued":{"date-parts":[[2026,10,5]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.07175","URL":"https://arxiv.org/abs/2610.07175","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.07175/","first_public":"2026-10-05","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_07591","type":"article","title":"Recurrent Looped Transformer","author":[{"literal":"Yifan Zhang"},{"literal":"Jichen Feng"},{"literal":"Shihan Qin"},{"literal":"Sun-Yuan Kung"},{"literal":"Mengdi Wang"}],"issued":{"date-parts":[[2026,10,6]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.07591","URL":"https://arxiv.org/abs/2610.07591","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.07591/","first_public":"2026-10-06","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_07730","type":"article","title":"SanSi: A Looped Typed Decision Model for System 1.5 Thinking","author":[{"literal":"Shuyu Gan"},{"literal":"Young-Jun Lee"},{"literal":"Dongyeop Kang"}],"issued":{"date-parts":[[2026,10,6]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.07730","URL":"https://arxiv.org/abs/2610.07730","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.07730/","first_public":"2026-10-06","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_07907","type":"article","title":"Continuous Memory Machines","author":[{"literal":"Ciaran Regan"},{"literal":"Kai Arulkumaran"},{"literal":"Luke Darlow"},{"literal":"Stefania Druga"},{"literal":"Sebastian Risi"},{"literal":"Llion Jones"}],"issued":{"date-parts":[[2026,10,6]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.07907","URL":"https://arxiv.org/abs/2610.07907","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.07907/","first_public":"2026-10-06","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_09145","type":"article","title":"Noise Your Prompt: Noising Conditioning Tokens in Continuous Diffusion Language Models","author":[{"literal":"Justin Jung"}],"issued":{"date-parts":[[2026,10,6]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.09145","URL":"https://arxiv.org/abs/2610.09145","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.09145/","first_public":"2026-10-06","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_09416","type":"article","title":"Efficient Reasoning with Flow Language Models","author":[{"literal":"Hanru Bai"},{"literal":"Faissal Izermine"},{"literal":"Oscar Davis"},{"literal":"T. Konstantin Rusch"}],"issued":{"date-parts":[[2026,10,7]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.09416","URL":"https://arxiv.org/abs/2610.09416","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.09416/","first_public":"2026-10-07","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_09876","type":"article","title":"Think Before You Paint: Recursive Latent Reasoning for Diffusion Models","author":[{"literal":"Paweł Skierś"},{"literal":"Małgorzata Grzanka"},{"literal":"Wojciech Masarczyk"},{"literal":"Jan-Willem van de Meent"},{"literal":"Kamil Deja"}],"issued":{"date-parts":[[2026,10,7]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.09876","URL":"https://arxiv.org/abs/2610.09876","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.09876/","first_public":"2026-10-07","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_10623","type":"article","title":"Recurrent Self-Improvement: Dynamic Cross-Loop On-Policy Distillation for Looped Language Models","author":[{"literal":"Yi Wang"},{"literal":"Rui Qian"},{"literal":"Yu Li"},{"literal":"Haoyang Yao"},{"literal":"Wenjie Wang"}],"issued":{"date-parts":[[2026,10,7]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.10623","URL":"https://arxiv.org/abs/2610.10623","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.10623/","first_public":"2026-10-07","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_10625","type":"article","title":"Safe at One Loop, Risky at Another: Aligning Safety Across Recurrent Depths in Looped Language Models","author":[{"literal":"Yi Wang"},{"literal":"Xiuyuan Qi"},{"literal":"Dongqi Han"},{"literal":"Dongsheng Li"},{"literal":"Wenjie Wang"}],"issued":{"date-parts":[[2026,10,7]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.10625","URL":"https://arxiv.org/abs/2610.10625","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.10625/","first_public":"2026-10-07","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_10996","type":"article","title":"Invariant-Measure Reasoners: Stable Representations for Latent Reasoning","author":[{"literal":"Yuto Inui"},{"literal":"Takuya Konishi"},{"literal":"Yoshinobu Kawahara"}],"issued":{"date-parts":[[2026,10,7]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.10996","URL":"https://arxiv.org/abs/2610.10996","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.10996/","first_public":"2026-10-07","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_20641","type":"article","title":"AbridgeKV: Compressing Long Reasoning into the KV Cache of Learned Pause Tokens","author":[{"literal":"Jack Cai"},{"literal":"Lijie Yang"},{"literal":"Vasilis Kontonis"},{"literal":"Dimitris Papailiopoulos"},{"literal":"Shivam Garg"}],"issued":{"date-parts":[[2026,10,7]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.20641","URL":"https://arxiv.org/abs/2610.20641","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.20641/","first_public":"2026-10-07","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_11570","type":"article","title":"Scaling to Tens of Thousands of Test-Time Iterations with Loop-Native Attention Residuals","author":[{"literal":"Pengxiang Li"},{"literal":"Dilxat Muhtar"},{"literal":"Di He"},{"literal":"Guinan Su"},{"literal":"Lu Yin"},{"literal":"Shiwei Liu"}],"issued":{"date-parts":[[2026,10,8]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.11570","URL":"https://arxiv.org/abs/2610.11570","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.11570/","first_public":"2026-10-08","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_12582","type":"article","title":"One Model to Loop Them All: Elastic Latent Recurrence for Language Models","author":[{"literal":"Faisal Ladhak"},{"literal":"Shantanu Acharya"},{"literal":"Santiago Akle Serano"},{"literal":"Simeng Sun"},{"literal":"Boris Ginsburg"},{"literal":"Krishna C Puvvada"}],"issued":{"date-parts":[[2026,10,8]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.12582","URL":"https://arxiv.org/abs/2610.12582","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.12582/","first_public":"2026-10-08","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_13330","type":"article","title":"Masked Diffusion Language Agents for Tool-Integrated Chemical Reasoning","author":[{"literal":"Mengdi Liu"},{"literal":"Chenghao Jia"},{"literal":"Hong Chang"},{"literal":"Shiguang Shan"}],"issued":{"date-parts":[[2026,10,9]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.13330","URL":"https://arxiv.org/abs/2610.13330","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.13330/","first_public":"2026-10-09","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_13527","type":"article","title":"S-Space: Exploring Spatial Workspace in Multimodal Language Models","author":[{"literal":"Zhiyuan Shi"},{"literal":"Xiang Li"},{"literal":"Xingzhuo Guo"},{"literal":"Shangchen Miao"},{"literal":"Weiliang Chen"},{"literal":"Qiyu Dai"},{"literal":"Jinshan Ren"},{"literal":"Jiangran Lyu"},{"literal":"Fangfu Liu"},{"literal":"Mingsheng Long"},{"literal":"Jialong Wu"}],"issued":{"date-parts":[[2026,10,9]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.13527","URL":"https://arxiv.org/abs/2610.13527","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.13527/","first_public":"2026-10-09","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_14252","type":"article","title":"Adaptive-Depth Is a Finishing Stage","author":[{"literal":"Andrei Piterbarg"},{"literal":"Sean McLeish"},{"literal":"Dhruv Pai"},{"literal":"Benjamin Keigwin"}],"issued":{"date-parts":[[2026,10,9]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.14252","URL":"https://arxiv.org/abs/2610.14252","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.14252/","first_public":"2026-10-09","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_14456","type":"article","title":"Learning to Extrapolate in Depth with Looped Transformers","author":[{"literal":"Huitao Yang"},{"literal":"Jianhao Huang"},{"literal":"Zixuan Wang"},{"literal":"Baharan Mirzasoleiman"}],"issued":{"date-parts":[[2026,10,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.14456","URL":"https://arxiv.org/abs/2610.14456","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.14456/","first_public":"2026-10-10","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_15081","type":"article","title":"Knowledge transfer between simple discrete worlds","author":[{"literal":"Silvère Gangloff"},{"literal":"Jan Hula"}],"issued":{"date-parts":[[2026,10,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.15081","URL":"https://arxiv.org/abs/2610.15081","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.15081/","first_public":"2026-10-10","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_15097","type":"article","title":"Anchors Enhance Controllability and Reasoning of Diffusion Language Models","author":[{"literal":"Feiyu Qu"},{"literal":"Zheng Huang"},{"literal":"Andrea Zanette"}],"issued":{"date-parts":[[2026,10,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.15097","URL":"https://arxiv.org/abs/2610.15097","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.15097/","first_public":"2026-10-10","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_15218","type":"article","title":"Running Fewer Loops Removes Refusal Behaviour in Recurrent-Depth Language Models","author":[{"literal":"Seyedarmin Azizi"},{"literal":"Erfan Baghaei Potraghloo"},{"literal":"Arya Fayyazi"},{"literal":"Saeid Shokoufa"},{"literal":"Massoud Pedram"}],"issued":{"date-parts":[[2026,10,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.15218","URL":"https://arxiv.org/abs/2610.15218","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.15218/","first_public":"2026-10-10","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_15501","type":"article","title":"Large Decision Models: Repurposing Large Language Models as Single-Pass Decision Makers","author":[{"literal":"Ziyi Chen"},{"literal":"Yan Zhang"},{"literal":"Jianhui Wei"},{"literal":"Jianing Yin"},{"literal":"Daoan Zhang"}],"issued":{"date-parts":[[2026,10,11]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.15501","URL":"https://arxiv.org/abs/2610.15501","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.15501/","first_public":"2026-10-11","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_16427","type":"article","title":"Resample the Answer: Where to Spend Test-Time Compute in Recursive Reasoning","author":[{"literal":"Maxime Illouz"}],"issued":{"date-parts":[[2026,10,12]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.16427","URL":"https://arxiv.org/abs/2610.16427","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.16427/","first_public":"2026-10-12","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_17679","type":"article","title":"Dynamic K-Head: Difficulty-Supervised Budget Selection for Continuous Latent Reasoning","author":[{"literal":"Ying Li"},{"literal":"Yuqi Chen"},{"literal":"Sixuan Li"},{"literal":"Yunfeng Cai"}],"issued":{"date-parts":[[2026,10,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.17679","URL":"https://arxiv.org/abs/2610.17679","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.17679/","first_public":"2026-10-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_18127","type":"article","title":"Monitoring latent reasoning in looped language models","author":[{"literal":"Pawel Olszowiec"},{"literal":"Paulina Tomaszewska"},{"literal":"Maciej Żelaszczyk"},{"literal":"Michal Byra"},{"literal":"Neo Christopher Chung"}],"issued":{"date-parts":[[2026,10,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.18127","URL":"https://arxiv.org/abs/2610.18127","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.18127/","first_public":"2026-10-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_18162","type":"article","title":"LoOPSD: On-Policy Self-Distillation with Looped Teachers","author":[{"literal":"Wei Fan"},{"literal":"Qingyue Yang"},{"literal":"Chao-Hong Tan"},{"literal":"Qian Chen"},{"literal":"Wen Wang"},{"literal":"Xiangang Li"},{"literal":"Kejiang Chen"},{"literal":"Weiming Zhang"},{"literal":"Nenghai Yu"}],"issued":{"date-parts":[[2026,10,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.18162","URL":"https://arxiv.org/abs/2610.18162","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.18162/","first_public":"2026-10-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_18174","type":"article","title":"Looped Reasoning Finishes Earlier Than You Think","author":[{"literal":"Sumin Park"},{"literal":"T. Konstantin Rusch"}],"issued":{"date-parts":[[2026,10,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.18174","URL":"https://arxiv.org/abs/2610.18174","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.18174/","first_public":"2026-10-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_18547","type":"article","title":"Input-Conditioned Search for Efficient Test-Time Scaling of Recursive Reasoners","author":[{"literal":"Haochen Shi"},{"literal":"Tianwen Fu"},{"literal":"Qianggang Ding"},{"literal":"Yinhan Lu"},{"literal":"Yizhan Li"},{"literal":"Huan Zhang"},{"literal":"Yajie Zhao"},{"literal":"Xingdi Yuan"},{"literal":"Hengyang Lu"},{"literal":"Bang Liu"}],"issued":{"date-parts":[[2026,10,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.18547","URL":"https://arxiv.org/abs/2610.18547","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.18547/","first_public":"2026-10-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_19080","type":"article","title":"Backpropagating Verifier Rewards Through Continuous Flow Language Models","author":[{"literal":"Xu Han"},{"literal":"Jiajing Hu"},{"literal":"Li-Ping Liu"}],"issued":{"date-parts":[[2026,10,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.19080","URL":"https://arxiv.org/abs/2610.19080","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.19080/","first_public":"2026-10-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_19220","type":"article","title":"Riding the Fixed-Point Curve: Image Classification in the Continuous Thought Machine","author":[{"literal":"Thomas Hannagan"}],"issued":{"date-parts":[[2026,10,13]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.19220","URL":"https://arxiv.org/abs/2610.19220","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.19220/","first_public":"2026-10-13","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_19408","type":"article","title":"ForkGRPO: Efficient Exploration and Fine-Grained Credit Assignment for Diffusion Language Models","author":[{"literal":"Zhaoyan Jia"},{"literal":"Chengwei Zhou"},{"literal":"Gourav Datta"},{"literal":"Massoud Pedram"}],"issued":{"date-parts":[[2026,10,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.19408","URL":"https://arxiv.org/abs/2610.19408","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.19408/","first_public":"2026-10-14","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_19568","type":"article","title":"INTACT: Inducing Latent Reasoning on a Frozen Backbone","author":[{"literal":"Kutay Tire"},{"literal":"Atula Tejaswi"},{"literal":"Haris Vikalo"},{"literal":"Sujay Sanghavi"}],"issued":{"date-parts":[[2026,10,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.19568","URL":"https://arxiv.org/abs/2610.19568","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.19568/","first_public":"2026-10-14","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_19768","type":"article","title":"Learning to Compute Is Not Learning to Retain or Reuse in Looped Transformers","author":[{"literal":"Peilong Zhou"},{"literal":"Hongtao Zhang"},{"literal":"Yulong Huang"},{"literal":"Cangyuan Li"},{"literal":"Mingyu Chen"},{"literal":"Ying Wang"}],"issued":{"date-parts":[[2026,10,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.19768","URL":"https://arxiv.org/abs/2610.19768","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.19768/","first_public":"2026-10-14","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_19871","type":"article","title":"Rationale Distribution Distillation via Flow Matching for Distributional Continuous Latent Reasoning","author":[{"literal":"Mingyeong Kim"},{"literal":"Gyusang Cho"},{"literal":"Juho Lee"},{"literal":"Giung Nam"}],"issued":{"date-parts":[[2026,10,14]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.19871","URL":"https://arxiv.org/abs/2610.19871","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.19871/","first_public":"2026-10-14","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_21182","type":"article","title":"Does Latent Chain-of-Thought Actually Reason? A Single, Unifying Taxonomy for Latent Reasoning","author":[{"literal":"Jingjie Qiu"},{"literal":"Sheldon Yu"},{"literal":"Xunyi Jiang"},{"literal":"Jiajin Liu"},{"literal":"Zihan Huang"},{"literal":"Zhendong Chu"},{"literal":"Tong Yu"},{"literal":"Julian McAuley"},{"literal":"Junda Wu"}],"issued":{"date-parts":[[2026,10,15]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.21182","URL":"https://arxiv.org/abs/2610.21182","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.21182/","first_public":"2026-10-15","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_2610_21594","type":"article","title":"LoopAhead: Speculating Ahead for Faster Decoding in Looped Language Models","author":[{"literal":"Zixuan Liu"},{"literal":"Nan Xue"},{"literal":"Rui Bao"},{"literal":"Yuhang Xu"},{"literal":"Xiaofeng Cao"},{"literal":"Jiangchao Yao"},{"literal":"Zhiyong Chen"}],"issued":{"date-parts":[[2026,10,15]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2610.21594","URL":"https://arxiv.org/abs/2610.21594","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2610.21594/","first_public":"2026-10-15","collection":"core","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},
{"id":"lit_precursor_1603_08983","type":"article","title":"Adaptive Computation Time for Recurrent Neural Networks","author":[{"literal":"Alex Graves"}],"issued":{"date-parts":[[2016,3,29]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:1603.08983","URL":"https://arxiv.org/abs/1603.08983","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/1603.08983/","first_public":"2016-03-29","collection":"precursor","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_precursor_1807_03819","type":"article","title":"Universal Transformers","author":[{"literal":"Mostafa Dehghani"},{"literal":"Stephan Gouws"},{"literal":"Oriol Vinyals"},{"literal":"Jakob Uszkoreit"},{"literal":"Łukasz Kaiser"}],"issued":{"date-parts":[[2018,7,10]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:1807.03819","URL":"https://arxiv.org/abs/1807.03819","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/1807.03819/","first_public":"2018-07-10","collection":"precursor","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_precursor_1909_01377","type":"article","title":"Deep Equilibrium Models","author":[{"literal":"Shaojie Bai"},{"literal":"J. Zico Kolter"},{"literal":"Vladlen Koltun"}],"issued":{"date-parts":[[2019,9,3]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:1909.01377","URL":"https://arxiv.org/abs/1909.01377","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/1909.01377/","first_public":"2019-09-03","collection":"precursor","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_precursor_2106_04537","type":"article","title":"Can You Learn an Algorithm? Generalizing from Easy to Hard Problems with Recurrent Networks","author":[{"literal":"Avi Schwarzschild"},{"literal":"Eitan Borgnia"},{"literal":"Arjun Gupta"},{"literal":"Furong Huang"},{"literal":"Uzi Vishkin"},{"literal":"Micah Goldblum"},{"literal":"Tom Goldstein"}],"issued":{"date-parts":[[2021,6,8]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2106.04537","URL":"https://arxiv.org/abs/2106.04537","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2106.04537/","first_public":"2021-06-08","collection":"precursor","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_precursor_2107_05407","type":"article","title":"PonderNet: Learning to Ponder","author":[{"literal":"Andrea Banino"},{"literal":"Jan Balaguer"},{"literal":"Charles Blundell"}],"issued":{"date-parts":[[2021,7,12]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2107.05407","URL":"https://arxiv.org/abs/2107.05407","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2107.05407/","first_public":"2021-07-12","collection":"precursor","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_precursor_2305_14699","type":"article","title":"Can Transformers Learn to Solve Problems Recursively?","author":[{"literal":"Shizhuo Dylan Zhang"},{"literal":"Curt Tigges"},{"literal":"Stella Biderman"},{"literal":"Maxim Raginsky"},{"literal":"Talia Ringer"}],"issued":{"date-parts":[[2023,5,24]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2305.14699","URL":"https://arxiv.org/abs/2305.14699","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2305.14699/","first_public":"2023-05-24","collection":"precursor","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_precursor_2311_12424","type":"article","title":"Looped Transformers are Better at Learning Learning Algorithms","author":[{"literal":"Liu Yang"},{"literal":"Kangwook Lee"},{"literal":"Robert Nowak"},{"literal":"Dimitris Papailiopoulos"}],"issued":{"date-parts":[[2023,11,21]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2311.12424","URL":"https://arxiv.org/abs/2311.12424","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2311.12424/","first_public":"2023-11-21","collection":"precursor","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}},{"id":"lit_precursor_2402_13572","type":"article","title":"AlgoFormer: An Efficient Transformer Framework with Algorithmic Structures","author":[{"literal":"Yihang Gao"},{"literal":"Chuanyang Zheng"},{"literal":"Enze Xie"},{"literal":"Han Shi"},{"literal":"Tianyang Hu"},{"literal":"Yu Li"},{"literal":"Michael K. Ng"},{"literal":"Zhenguo Li"},{"literal":"Zhaoqiang Liu"}],"issued":{"date-parts":[[2024,2,21]]},"genre":"Preprint","archive":"arXiv","number":"arXiv:2402.13572","URL":"https://arxiv.org/abs/2402.13572","custom":{"record_url":"https://latent-implicit-thinking.github.io/papers/2402.13572/","first_public":"2024-02-21","collection":"precursor","citation_basis":"ArXiv preprint, using its initial submission date; source URL is not version-pinned."}}
]