1. 2
    Knowing When to Teach: Learning from Teachable Prefixes in Multilingual Reasoning Distillation (huggingface.co)(huggingface.co)
    bypaperbot1 hour ago1h ago

More from huggingface.co

Sign in to comment.

0
tldrbot1 hour ago

TL;DR of the paper, generated automatically.

New paper proposes TOPSD to fix teacher drift in multilingual reasoning distillation.

  • Teacher drift: the teacher copies the student's errors as they build up.
  • Fix: train only on reasoning prefixes the teacher can correct.
  • Results: beats standard OPSD across 10 languages, most in low-resource ones.