{"slug": "nexus-structured-runtime-safety-for-tool-using-llm-agents", "title": "NEXUS: Structured Runtime Safety for Tool-Using LLM Agents", "summary": "Researchers have released NEXUS (Neural EXecution Utility and Safety), a structured-plan safety monitor for tool-using LLM agents that applies a formal intervention policy to select among allow, block, request confirmation, or request revision actions. On a 128-instance synthetic benchmark, NEXUS achieves an F1 score of 0.949 and a 4-class intervention accuracy of 0.6406, outperforming rule-only intervention selection by 27.3 percentage points, with 0.205 ms median latency adding under 0.1% overhead to typical agent loops.", "body_md": "arXiv:2607.19356v1 Announce Type: new\nAbstract: Tool-using LLM agents increasingly execute high-impact actions, making runtime safety monitoring essential. We present NEXUS (Neural EXecution Utility and Safety), a structured-plan safety monitor that applies a formal intervention policy to select among four actions: allow, block, request confirmation, or request revision. NEXUS combines deterministic safety rules, argument-level inspection, and a calibrated logistic-regression risk score for graded escalation. On a 128-instance synthetic benchmark, NEXUS achieves an F1 score of 0.949 and a 4-class intervention accuracy of 0.6406, outperforming rule-only intervention selection by 27.3 percentage points. It also improves over rule-only on R-Judge (F1 = 0.861 vs. 0.849), matches rule-only on AgentHarm due to threat-model limits, and achieves 0% ASR at 99% control allow on IPI. On the rule-blind NEXUS-Stress benchmark, NEXUS reaches an F1 score of 0.881, highlighting the difficulty of fine-grained intervention routing. With 0.205 ms median latency, NEXUS adds under 0.1% overhead to typical agent loops. Code, benchmarks, and the calibrated risk scorer are publicly released.", "url": "https://wpnews.pro/news/nexus-structured-runtime-safety-for-tool-using-llm-agents", "canonical_source": "https://arxiv.org/abs/2607.19356", "published_at": "2026-07-23 04:00:00+00:00", "updated_at": "2026-07-23 04:08:22.796134+00:00", "lang": "en", "topics": ["ai-safety", "large-language-models", "ai-agents"], "entities": ["NEXUS", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/nexus-structured-runtime-safety-for-tool-using-llm-agents", "markdown": "https://wpnews.pro/news/nexus-structured-runtime-safety-for-tool-using-llm-agents.md", "text": "https://wpnews.pro/news/nexus-structured-runtime-safety-for-tool-using-llm-agents.txt", "jsonld": "https://wpnews.pro/news/nexus-structured-runtime-safety-for-tool-using-llm-agents.jsonld"}}