DART-SD：面向多轮工具调用智能体的菱形拓扑感知自蒸馏框架
原标题：Paper page - DART-SD: Diamond-topology Aware Retrieval and Tuning for Self-Distillation of Multi-Turn Tool-Calling Agents
AI 摘要
DART-SD 是一个用于多轮工具调用智能体的自蒸馏框架，它将执行过程建模为菱形拓扑图，识别关键拓扑断点，并仅对恢复步骤进行局部监督，以保留正确的推理前缀。实验表明，DART-SD 在五个工具使用基准和两个模型规模上优于传统的全轨迹基线，其中 Qwen3-8B 学生模型在多个基准上甚至超过了更大的教师模型。
正文节选
DART-SD: Diamond-topology Aware Retrieval and Tuning for Self-Distillation of Multi-Turn Tool-Calling Agents Abstract DART-SD improves multi-turn tool-calling agents by modeling execution as a diamond-topology graph, identifying critical failure points, and applying localized self-distillation to preserve valid reasoning while correcting errors. Equipping Large Language Models (LLMs) with multi-turn tool-calling capabilities is essential for building autonomous agents. However, progress is funda