5 papers
Rethinking On-Policy Self-Distillation for Thinking Models
Simran Kaur, Narutatsu Ri, Yinghui He +2
Self-distillation is a promising recipe for self-improvement in language models. In this setting, a model can serve as its own teacher when given privileged information, such as a…
Self-Distillation Zero: Self-Revision Turns Binary Rewards into Dense Supervision
Yinghui He, Simran Kaur, Adithya Bhaskar +7
Current post-training methods in verifiable settings fall into two categories. Reinforcement learning (RLVR) relies on binary rewards, which are broadly applicable and powerful, bu…
How Does RL Post-training Induce Skill Composition? A Case Study on Countdown
Simon Park, Simran Kaur, Sanjeev Arora
While reinforcement learning (RL) successfully enhances reasoning in large language models, its role in fostering compositional generalization (the ability to synthesize novel skil…
Instruct-SkillMix: A Powerful Pipeline for LLM Instruction Tuning
Simran Kaur, Simon Park, Anirudh Goyal +1
We introduce Instruct-SkillMix, an automated approach for creating diverse, high quality SFT data for instruction-following. The pipeline involves two stages, each leveraging an ex…
Can Models Learn Skill Composition from Examples?
Haoyu Zhao, Simran Kaur, Dingli Yu +2
As large language models (LLMs) become increasingly advanced, their ability to exhibit compositional generalization -- the capacity to combine learned skills in novel ways not enco…