Researchers propose ECOD, an experience-calibrated on-policy distillation method for multi-turn agents
A new arXiv paper introduces ECOD, Experience-Calibrated On-Policy Distillation, targeting the challenge of sparse and delayed rewards in reinforcement learning for long-horizon language agents, complementing on-policy distillation.
Read the full story at arXiv cs.LG ↗
Timeline · 1 report
- 2026-11-17 05:00 · arXiv cs.LG
ECOD: Experience-Calibrated On-Policy Distillation for Multi-Turn Agents