SWE-OPD distills agentic coding from teacher into smaller models
AI-generated summary Researchers present SWE-OPD, a study of on-policy distillation for transferring agentic software engineering capability from a large teacher model to a smaller student model. Supervision from Qwen3.6-27B lifts the SWE-bench Pro resolve rate of Qwen3.5-9B from 35.9% to 39.4%.