Researchers have explored On-Policy Delta Distillation (OPD^2), an advancement over On-Policy Distillation (OPD), for multilingual mathematical reasoning. Experiments using the Qwen3 model demonstrated that OPD^2 significantly outperforms standard OPD across English, Korean, and Japanese. The study also noted that while English-only OPD can boost performance in other languages, it may inadvertently shift responses towards English, underscoring the need for multilingual data to maintain language-specific responses. AI
IMPACT This research could lead to more effective multilingual LLMs for complex reasoning tasks.
RANK_REASON The cluster contains a research paper detailing a new technique for LLM post-training. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
