Back to today's list

NEXUS: Structured Runtime Safety for Tool-Using LLM Agents

Elias Hossain, Md Mehedi Hasan Nipu, Tasfia Nuzhat Ornee, Rajib Rana, Niloofar Yousefi

Published Jul 23, 2026Featured #3In the daily list Jul 24, 2026
Daily score67.4
Editorial review7.2
Relevance0.466
Freshness0.722

Why It Matters

What makes this one worth your time

As LLM agents increasingly perform high-impact actions, ensuring their safe operation is crucial for preventing unintended consequences in real-world applications.

NEXUS enhances runtime safety for LLM agents with structured intervention policies.

Summary

The paper introduces NEXUS, a structured runtime safety monitor for tool-using LLM agents, which uses a combination of deterministic safety rules, argument-level inspection, and a logistic-regression risk score to manage interventions during agent execution. It demonstrates improved performance over rule-only systems on several benchmarks and maintains low latency.

Key contributions

  • Development of NEXUS, a structured runtime safety monitor for LLM agents.
  • Integration of deterministic safety rules with a logistic-regression risk score for intervention management.
  • Public release of code, benchmarks, and a calibrated risk scorer.

Notable insights

  • Combining deterministic rules with a logistic-regression risk score allows for graded intervention escalation.
  • The approach achieves significant performance improvements over rule-only systems on specific benchmarks.

Possible limitations

  • Not stated in the abstract

Abstract

arXiv:2607.19356v1 Announce Type: new Abstract: 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.