arXiv cs.AI· Aniruddha Chattopadhyay, Kanishka Halder, Partha Sarathi Roy·· 12 小时前AI 评分35
SchemaShield-Lite：无需重训的 LLM 智能体函数调用接口漂移修复层
Robust Function Calling for LLM Agents Under Interface Drift
AI 导读
SchemaShield-Lite 是一种无需训练的修复层，用于应对 LLM 智能体部署后工具描述、参数 schema 与候选工具列表变化引发的接口漂移问题。在 Qwen3.5-9B 上，它将 DICE-300 准确率从 0.717 提升至 0.813，BFCL-200 从 0.580 提升至 0.805，分别挽回 19/22 和 33/45 的漂移致失败且未观察到修复损害。
正文 · 原文
Abstract:Tool-augmented language-model agents rely on interface text: tool descriptions, argument schemas, and the candidate tool list shown at inference time. These interfaces often change after deployment even when the underlying tool semantics remain fixed, creating a practical reliability problem for function-calling systems. We study this failure mode as interface drift and evaluate whether a lightweight, training-free repair layer can recover robustness without retraining the base model. Our method, SchemaShield-Lite, combines schema validation, candidate-set retry for missing tool names, and one-shot constrained repair. On Qwen3.5-9B, it raises DICE-300 accuracy from 0.717 to 0.813 and BFCL-200 accuracy from 0.580 to 0.805, recovering 19/22 and 33/45 drift-induced failures with no observed repair harms. Naive retry does not explain the gains. Ablations show that candidate-set drift, especially distractor tools, accounts for most observed damage, while description and schema edits alone are less harmful. The results position validation-and-repair as a practical robustness layer for deployed LLM-agent tool use.
| Comments: | 16 pages, 1 figure, 6 tables. Peer-reviewed and accepted at NGEN-AI 2026 (International Conference on Next Generation AI Systems). This is the pre-review submitted version |
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Software Engineering (cs.SE) |
| Cite as: | arXiv:2610.23342 [cs.AI] |
| (or arXiv:2610.23342v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.23342 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Aniruddha Chattopadhyay [view email]
[v1]
Fri, 16 Oct 2026 11:49:37 UTC (35 KB)
来源：arXiv cs.AI · arxiv.org