2 papers
cs.CL2025
Through the Valley: Path to Effective Long CoT Training for Small Language Models
Renjie Luo, Jiaxi Li, Chen Huang +1
Long chain-of-thought (CoT) supervision has become a common strategy to enhance reasoning in language models. While effective for large models, we identify a phenomenon we call Lon…
cs.CL2025
Language Models Can Learn from Verbal Feedback Without Scalar Rewards
Renjie Luo, Zichen Liu, Xiangyan Liu +5
LLMs are often trained with RL from human or AI feedback, yet such methods typically compress nuanced feedback into scalar rewards, discarding much of their richness and inducing s…