LIT Workshop at ICLR 2026
Latent & Implicit Thinking – Going Beyond CoT Reasoning
The Latent & Implicit Thinking – Going Beyond CoT Reasoning (LIT) workshop was held at ICLR 2026. While chain-of-thought reasoning aligns intuitively with human cognition, the workshop asked whether language is the most effective internal computation medium for neural networks.
The program brought together research on continuous hidden representations, discrete reasoning tokens, looped architectures, KV-cache and hybrid stateful reasoning, mechanistic interpretability, non-autoregressive approaches, theory, evaluation, limitations, and safety.
The workshop site now also hosts the living Implicit Chain-of-Thought paper map, with a searchable chronology, method guides, and machine-readable data.
Workshop archive
Original announcement
The LIT workshop aimed to bring together researchers across the spectrum of implicit reasoning to spark collaboration and accelerate progress in this emerging area.
The workshop took place on April 27, 2026, in Rio de Janeiro, following the ICLR 2026 main conference.