Wei Lu

PEAR: Phase Entropy Aware Reward for Efficient Reasoning

Difficulty–Diversity Collaborative Filtering for Data-Efficient LLM Fine-Tuning

Language Models Can Learn from Verbal Feedback Without Scalar Rewards

WildLong: Synthesizing Realistic Long-Context Instruction Data at Scale

EvoScale: Evolutionary Test-Time Scaling for Software Engineering

Satori: Reinforcement Learning with Chain-of-Action-Thought Enhances LLM Reasoning via Autoregressive Search

Synthetic Data (Almost) from Scratch: Generalized Instruction Tuning for Language Models