{"slug": "robust-function-calling-for-llm-agents-under-interface-drift", "title": "Robust Function Calling for LLM Agents Under Interface Drift", "summary": "A training-free repair layer called SchemaShield-Lite raised function-calling accuracy for tool-augmented LLM agents under interface drift, according to an arXiv paper (2610.23342v1). On Qwen3.5-9B, SchemaShield-Lite lifted DICE-300 accuracy from 0.717 to 0.813 and BFCL-200 accuracy from 0.580 to 0.805, recovering 19 of 22 and 33 of 45 drift-induced failures respectively with no observed repair harms. The authors report that candidate-set drift, especially distractor tools, accounts for most of the damage, while description and schema edits alone are less harmful, positioning validation-and-repair as a practical robustness layer for deployed LLM-agent tool use.", "body_md": "arXiv:2610.23342v1 Announce Type: new \nAbstract: 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.", "url": "https://wpnews.pro/news/robust-function-calling-for-llm-agents-under-interface-drift", "canonical_source": "https://www.machinebrief.com/news/robust-function-calling-for-llm-agents-under-interface-drift-925m", "published_at": "2026-10-19 04:00:00+00:00", "updated_at": "2026-10-19 04:50:45.929406+00:00", "lang": "en", "topics": ["ai-agents", "large-language-models", "ai-research", "ai-tools"], "entities": ["SchemaShield-Lite", "Qwen3.5-9B", "DICE-300", "BFCL-200", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/robust-function-calling-for-llm-agents-under-interface-drift", "markdown": "https://wpnews.pro/news/robust-function-calling-for-llm-agents-under-interface-drift.md", "text": "https://wpnews.pro/news/robust-function-calling-for-llm-agents-under-interface-drift.txt", "jsonld": "https://wpnews.pro/news/robust-function-calling-for-llm-agents-under-interface-drift.jsonld"}}