Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning to Rewrite Tool Descriptions for Reliable LLM-Agent Tool Use

Ruocheng Guo, Kaiwen Dong, Xiang Gao, Kamalika Das

TL;DR

Trace-Free+, a curriculum learning framework that progressively transfers supervision from trace-rich settings to trace-free deployment, encouraging the model to abstract reusable interface-usage patterns and tool usage outcomes, is proposed.

Abstract

The performance of LLM-based agents depends not only on the agent itself but also on the quality of the tool interfaces it consumes. While prior work has focused heavily on agent fine-tuning, tool interfaces-including natural language descriptions and parameter schemas-remain largely human-oriented and often become a bottleneck, especially when agents must select from large candidate tool sets. Existing approaches to improving tool interfaces rely on execution traces, which are frequently unavailable in cold-start or privacy-constrained settings, and typically optimize each tool independently, limiting scalability and generalization to unseen tools. We propose Trace-Free+, a curriculum learning framework that progressively transfers supervision from trace-rich settings to trace-free deployment, encouraging the model to abstract reusable interface-usage patterns and tool usage outcomes. To support this approach, we construct a large-scale dataset of high-quality tool interfaces using a structured workflow over a diverse collection of tools. Experiments on StableToolBench and RestBench show consistent gains on unseen tools, strong cross-domain generalization, and robustness as the number of candidate tools scales to over 100, demonstrating that tool interface optimization is a practical and deployable complement to agent fine-tuning.

Learning to Rewrite Tool Descriptions for Reliable LLM-Agent Tool Use

TL;DR

Trace-Free+, a curriculum learning framework that progressively transfers supervision from trace-rich settings to trace-free deployment, encouraging the model to abstract reusable interface-usage patterns and tool usage outcomes, is proposed.

Abstract

The performance of LLM-based agents depends not only on the agent itself but also on the quality of the tool interfaces it consumes. While prior work has focused heavily on agent fine-tuning, tool interfaces-including natural language descriptions and parameter schemas-remain largely human-oriented and often become a bottleneck, especially when agents must select from large candidate tool sets. Existing approaches to improving tool interfaces rely on execution traces, which are frequently unavailable in cold-start or privacy-constrained settings, and typically optimize each tool independently, limiting scalability and generalization to unseen tools. We propose Trace-Free+, a curriculum learning framework that progressively transfers supervision from trace-rich settings to trace-free deployment, encouraging the model to abstract reusable interface-usage patterns and tool usage outcomes. To support this approach, we construct a large-scale dataset of high-quality tool interfaces using a structured workflow over a diverse collection of tools. Experiments on StableToolBench and RestBench show consistent gains on unseen tools, strong cross-domain generalization, and robustness as the number of candidate tools scales to over 100, demonstrating that tool interface optimization is a practical and deployable complement to agent fine-tuning.
Paper Structure (40 sections, 3 figures, 9 tables)

This paper contains 40 sections, 3 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Problem Statement
  3. Methodology
  4. Data Synthesis for Tool Interface Improvement
  5. Agentic Seed Tool Annotation and Filtering
  6. User Query Synthesis
  7. Workflow for Trace-based Tool Interface Improvement
  8. Trace-free+: Training Generalizable Tool Description Generators
  9. Inference for Unseen Tools
  10. Experiments
  11. Experimental Setup
  12. Evaluation Protocol.
  13. Evaluation Metrics.
  14. Data Pre-processing.
  15. Trace-free Evaluation
  16. ...and 25 more sections

Figures (3)

  • Figure 1: An illustration of the proposed tool interface improvement pipeline. Compared to the original description ($D0$), the learned description generator produces more effective tool descriptions that lead to better tool usage.
  • Figure 2: The SFT data synthesis pipeline.
  • Figure 3: Scaling experiment results for the trace-free setting on G2 Category, G2 Instruction and G3 Instruction subsets of StableToolBench. We report query-level (QL) success rate.