๐ LIT Workshop @ ICLR 2026 โ Community Choice Award!
Vote for your favorite paper from our Best Paper finalists ๐
Details on each paper in the thread ๐งต
4๏ธโฃ When does CoT Help? (https://t.co/LQSoJxHzln) @YijunDong1
Models chain-of-thought as a Markov chain and identifies "transition alignment" as the key factor determining when CoT actually helps. Theory + synthetic benchmarks to validate predictions.
๐ LIT Workshop @ ICLR 2026 โ Community Choice Award!
Vote for your favorite paper from our Best Paper finalists ๐
Details on each paper in the thread ๐งต
3๏ธโฃ Think-at-Hard (https://t.co/SK3dpo8Bx6)
Tackles latent overthinking in looped transformers โ a neural decider triggers extra iterations only on hard tokens. Skips 93% of tokens while outperforming always-iterate baselines by 3.8โ4.4%.