Yun Luo

Value Flattening is identified as an important yet overlooked failure mode of critic learning in standard PPO and a simple sparse supervision strategy can mitigate it; SParse Proximal Policy Optimization is introduced, which applies the value loss to only a few well-separated states in each response to mitigate both effects.

Short-context models can achieve superior reasoning performance by leveraging long-context teacher models through innovative token alignment and training strategies.

WIFA reduces harmful refusal while minimizing benign over-refusal, achieving a remarkable drop in over-refusal rates from 25.7% to 17.4%.

Even the best LLMs struggle with Olympiad-level combinatorics, achieving only 65.4% on a benchmark designed to expose their reasoning limitations.

Draft-OPD accelerates inference by over 5x while improving speculative decoding accuracy, transforming how draft models learn from target feedback.

Test-time training can finally scale for large reasoning models: TEMPO unlocks sustained performance gains by interleaving policy refinement with periodic critic recalibration, boosting accuracy by over 18% on challenging benchmarks.

LLMs can be taught to "think longer" and explore more diverse reasoning paths in-context via a simple length-incentivized reward, leading to improved generalization.