TopoCurate:Modeling Interaction Topology for Tool-Use Agent Training

2026-03-02

AI Summary

This work proposes TopoCurate, a novel framework that addresses the limitations of existing agent-training approaches which rely primarily on final success rates to filter trajectories, thereby overlooking the dynamic structure of interaction processes and conflating effective strategies with accidental successes. TopoCurate introduces semantic quotient topology modeling to map multi-turn trajectories onto a structured manifold, explicitly capturing strategy bifurcations induced by tool invocations through semantic-equivalence state merging and error-recovery path identification. Building on this representation, the framework implements a dual curation mechanism tailored for both supervised fine-tuning (SFT) and reinforcement learning (RL), enabling structure-aware evaluation of trajectory quality and informativeness. Evaluated on BFCLv3 and Tau2 Bench, TopoCurate achieves performance gains of 4.2% and 6.9% over current state-of-the-art baselines in SFT and RL settings, respectively, significantly enhancing agent generalization and robustness.

Technology Category

• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Multiagent Systems: Multiagent Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Training tool-use agents typically relies on outcome-based filtering: Supervised Fine-Tuning (SFT) on successful trajectories and Reinforcement Learning (RL) on pass-rate-selected tasks. However, this paradigm ignores interaction dynamics: successful trajectories may lack error recovery or exhibit redundancy, while pass rates fail to distinguish structurally informative tasks from trivial ones. We propose \textbf{TopoCurate}, an interaction-aware framework that projects multi-trial rollouts from the same task into a unified semantic quotient topology. By merging equivalent action-observation states, this projection transforms scattered linear trajectories into a structured manifold that explicitly captures how tool invocations and environmental responses drive the divergence between effective strategies and failure modes. Leveraging this representation, we introduce a dual-selection mechanism: for SFT, we prioritize trajectories demonstrating reflective recovery, semantic efficiency, and strategic diversity to mitigate covariate shift and mode collapse; for RL, we select tasks with high error branch ratios and strategic heterogeneity, maximizing gradient Signal-to-Noise Ratio to address vanishing signals in sparse-reward settings. Evaluations on BFCLv3 and Tau2 Bench show that TopoCurate achieves consistent gains of 4.2\% (SFT) and 6.9\% (RL) over state-of-the-art baselines. We will release the code and data soon for further investigations.

Problem

• tool-use agents
• interaction dynamics
• trajectory filtering
• task selection
• sparse-reward

Innovation

• interaction topology
• trajectory curation
• semantic quotient space
• dual-selection mechanism
• tool-use agents