ReNIO: Reweighting Negative Trajectory Importance for LLM On-Policy Distillation
ReNIO enhances on-policy distillation for language models by reweighting negative trajectories based on token-level probability ratios, improving reasoning performance in mathemati…
Hugging Face · Daily Papers
·Chen Lin, Kedi Chen
·
·▲ 5 upvotes
Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.
Autores: Chen Lin, Kedi Chen, Wei Zhang
- 5 upvotes da comunidade
- Temas: on-policy distillation, student model, teacher model, generated outputs, reasoning traces, exploration behavior
Resumo
Resumo original (em inglês), extraído do paper:
ReNIO enhances on-policy distillation for language models by reweighting negative trajectories based on token-level probability ratios, improving reasoning performance in mathematical and code generation tasks.
