DART-SD：多輪工具調用智能體自蒸餾的鑽石拓撲感知檢索與調優

DART-SD: Diamond-topology Aware Retrieval and Tuning for Self-Distillation of Multi-Turn Tool-Calling Agents

August 19, 2026

作者: Hangrui Xu, Jiarui Wang, Yang Yang, Chuanbo Zhu, Fangda Chen, Ziqi Wu, Jingming Cai, Yan Song

cs.AI

摘要

賦予大型語言模型（LLM）多輪工具調用能力，對於構建自主智慧體（autonomous agents）至關重要。然而，該領域的進展從根本上受到對全長軌跡模仿（full-length trajectory imitation）之依賴的限制。對於涉及多個順序無關之子目標的任務，最優解空間會形成一個龐大的組合鑽石晶格（combinatorial diamond lattice）。將這種豐富的拓撲結構強行壓縮為單調的軌跡，會導致嚴重的拓撲坍縮（topological collapse），不分青紅皂白地懲罰合理的替代性探索路徑，並嚴重削弱策略多樣性。為了解決這一問題，我們提出了 DART-SD（Diamond-topology Aware Retrieval and Tuning for Self-Distillation，基於鑽石拓撲感知的檢索與自蒸餾調優），這是一個新穎的框架，將範式從全局強制轉變為拓撲引導的局部修正。DART-SD 首先將執行過程建模為一個收斂的交互狀態轉移圖（Interaction-State Transition Graph, ISTG），忠實地捕捉成功與失敗探索路徑中所蘊含的鑽石拓撲結構。在自主滾動（autonomous rollouts）過程中，該框架會識別關鍵拓撲斷點（Critical Topological Breakpoint, CTB），並檢索由成功經驗支持的恢復參考。最後，我們通過 CTB 引導的局部監督引入了一種漸進式自蒸餾範式，確保訓練損失僅在生成的恢復步驟上計算，同時嚴格保護有效的推理前綴免受破壞性梯度更新的影響。在複雜多輪工具調用基準上的實驗表明，DART-SD 顯著優於傳統的全軌跡基線方法。

English

Equipping Large Language Models (LLMs) with multi-turn tool-calling capabilities is essential for building autonomous agents. However, progress is fundamentally limited by the reliance on full-length trajectory imitation. For tasks involving multiple order-independent sub-goals, the optimal solution space forms a vast combinatorial diamond lattice. Forcing this rich topology into monolithic trajectories causes a severe topological collapse, indiscriminately penalizing valid alternative explorations and severely degrading policy diversity. To address this, we propose DART-SD (Diamond-topology Aware Retrieval and Tuning for Self-Distillation), a novel framework that shifts the paradigm from global forcing to topology-guided localized correction. DART-SD first models the execution process as a converging Interaction-State Transition Graph (ISTG), faithfully capturing the inherent diamond topology of successful and failed exploratory paths. During autonomous rollouts, the framework identifies the Critical Topological Breakpoint (CTB) and retrieves success-supported recovery references. Finally, we introduce a progressive self-distillation paradigm through CTB-guided localized supervision, ensuring that the training loss is calculated exclusively on the generated recovery steps while strictly protecting the valid reasoning prefix from destructive gradient updates. Experiments on complex multi-turn tool-calling benchmarks demonstrate that DART-SD significantly outperforms traditional full-trajectory baselines.