AnalysisAI ModelsJuly 17, 2026

Multiple arxiv papers on on-policy distillation for LLMs

Four new arxiv papers (SEED, ShortOPD, Demystifying OPD, EasyOPD) explore on-policy distillation (OPD) for LLMs. SEED targets agentic RL, ShortOPD recovers pruned models, Demystifying OPD analyzes pathologies, and EasyOPD offers a unified framework.

4 sources