Paper Detail

On-Policy Delta Distillation for Multilingual Math Reasoning

Byeongho Heo, Jaehui Hwang, Sangdoo Yun, Dongyoon Han

huggingface Score 9.4

Published 2026-08-06 · First seen 2026-08-07

General AI

Abstract

On-Policy Distillation (OPD) is emerging as a promising alternative to reinforcement learning for LLM post-training, yet its effectiveness in multilingual settings remains underexplored. We study OPD and its advanced variant, On-Policy Delta Distillation (OPD^2), for mathematical reasoning in English, Korean, and Japanese. OPD^2 improves OPD by using the probability gap between a post-trained teacher and its base model as the learning signal. Experiments with Qwen3 show that OPD^2 consistently outperforms the original OPD, with particularly strong improvements in Korean and Japanese, and generally narrows the English-Korean performance gap. We further find that English-only OPD can also increase performance for Korean and Japanese, but often shifts the responses toward English, highlighting the importance of multilingual data to preserving target-language responses.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
soon
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@misc{heo2026policy,
  title = {On-Policy Delta Distillation for Multilingual Math Reasoning},
  author = {Byeongho Heo and Jaehui Hwang and Sangdoo Yun and Dongyoon Han},
  year = {2026},
  abstract = {On-Policy Distillation (OPD) is emerging as a promising alternative to reinforcement learning for LLM post-training, yet its effectiveness in multilingual settings remains underexplored. We study OPD and its advanced variant, On-Policy Delta Distillation (OPD\textasciicircum{}2), for mathematical reasoning in English, Korean, and Japanese. OPD\textasciicircum{}2 improves OPD by using the probability gap between a post-trained teacher and its base model as the learning signal. Experiments with Qwen3 show that OPD\textasciicircum{}2 consistently o},
  url = {https://huggingface.co/papers/2608.05802},
  keywords = {huggingface daily},
  eprint = {2608.05802},
  archiveprefix = {arXiv},
}

Metadata

{}