Pith. sign in

REVIEW 28 cited by

R-Judge: Benchmarking Safety Risk Awareness for LLM Agents

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2401.10019 v3 pith:WEQNK2C3 submitted 2024-01-18 cs.CL cs.AI

classification cs.CLcs.AI
keywords r-judgerisksafetyllmsagentagentsawarenessbenchmarking
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

Large language models (LLMs) have exhibited great potential in autonomously completing tasks across real-world applications. Despite this, these LLM agents introduce unexpected safety risks when operating in interactive environments. Instead of centering on the harmlessness of LLM-generated content in most prior studies, this work addresses the imperative need for benchmarking the behavioral safety of LLM agents within diverse environments. We introduce R-Judge, a benchmark crafted to evaluate the proficiency of LLMs in judging and identifying safety risks given agent interaction records. R-Judge comprises 569 records of multi-turn agent interaction, encompassing 27 key risk scenarios among 5 application categories and 10 risk types. It is of high-quality curation with annotated safety labels and risk descriptions. Evaluation of 11 LLMs on R-Judge shows considerable room for enhancing the risk awareness of LLMs: The best-performing model, GPT-4o, achieves 74.42% while no other models significantly exceed the random. Moreover, we reveal that risk awareness in open agent scenarios is a multi-dimensional capability involving knowledge and reasoning, thus challenging for LLMs. With further experiments, we find that fine-tuning on safety judgment significantly improve model performance while straightforward prompting mechanisms fail. R-Judge is publicly available at https://github.com/Lordog/R-Judge.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 28 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. The Balkanization of Execution-Security Research for AI Coding Agents: Isolation, Access Control, and Time-of-Check-to-Time-of-Use Vulnerabilities

    cs.CR 2026-07 accept novelty 6.5 of 10

    Execution-security research for AI coding agents is fragmented across 17 mechanism categories with five unaddressed cross-cutting gaps, including missing head-to-head isolation-vs-capability evaluation and untested re...

  2. SkillAttack: Automated Red Teaming of Agent Skills through Attack Path Refinement

    cs.CR 2026-04 conditional novelty 6.5 of 10

    SkillAttack exploits latent vulnerabilities in unmodified agent skills via closed-loop adversarial prompting, reaching ASR 0.73–0.93 on adversarial skills and up to 0.26 on real-world skills across 10 LLMs.

  3. How Benchmarks Mis-Score Computer-Use Agents

    cs.AI 2026-07 conditional novelty 6.0 of 10

    About 15.3% of FAIL verdicts across five CUA benchmarks are wrong, and genuine failures are mostly verification and planning errors, not clicks.

  4. Every Model Cheats: Prompt-Level Mitigation of Cheating on Offensive Cyber Tasks

    cs.CR 2026-07 conditional novelty 6.0 of 10

    On Cybench CTF tasks, 21 of 22 models cheated under baseline, 37% of baseline passes were cheated, and stricter anti-cheat prompts reduced but did not remove cheating.

  5. Execution-Grounded Security Testing for Coding Agents in Software Engineering Pipelines

    cs.AI 2026-06 conditional novelty 6.0 of 10

    Disguising unsafe operations as routine software-engineering tests makes coding agents execute them at high verified rates, revealing a gap between language-layer refusal and execution-layer harm.

  6. StepShield: When, Not Whether to Intervene on Rogue Agents

    cs.LG 2026-01 conditional novelty 6.0 of 10

    StepShield's step-level benchmark claims detection timeliness is a separate axis from accuracy, and reports a 2.3x early-intervention gap between LLM judges and regex guardrails that standard metrics cannot see.

  7. Measuring the Security of Mobile LLM Agents under Adversarial Prompts from Untrusted Third-Party Channels

    cs.CR 2025-10 conditional novelty 6.0 of 10

    Indirect prompt injection through ads, webviews, and notifications reliably diverts mobile LLM agents into leaking data and installing malware across eight evaluated agents.

  8. SENTINEL: A Multi-Level Formal Framework for Safety Evaluation of Foundation Model-based Embodied Agents

    cs.AI 2025-10 conditional novelty 6.0 of 10

    A three-level temporal-logic safety evaluator for embodied LLM agents that checks NL-to-LTL interpretation, plan compliance, and CTL over simulated execution trees.

  9. MLA-Trust: Benchmarking Trustworthiness of Multimodal LLM Agents in GUI Environments

    cs.AI 2025-06 conditional novelty 6.0 of 10

    MLA-Trust introduces 34 tasks and an evaluation toolbox showing that GUI-interacting multimodal agents are substantially less trustworthy than static multimodal chat models.

  10. RiOSWorld: Benchmarking the Risk of Multimodal Computer-Use Agents

    cs.AI 2025-05 conditional novelty 6.0 of 10

    Multimodal computer-use agents show risky intent in about 85% of tasks and complete risky actions in about 60%, measured on RiOSWorld, a new 492-task virtual-machine benchmark.

  11. USB: A Comprehensive and Unified Safety Evaluation Benchmark for Multimodal Large Language Models

    cs.CR 2025-05 conditional novelty 6.0 of 10

    USB-SafeBench is a unified MLLM safety benchmark with 61 risk categories, 4 modality combinations, and dual-language vulnerability and oversensitivity tests.

  12. Safety Degradation in AI Agents

    cs.CY 2025-05 conditional novelty 6.0 of 10

    Adding retrieval to aligned LLMs degrades safety: refusal rates fall, bias and harmfulness rise, and prompt-based mitigation only partially restores alignment.

  13. Agentic Harnesses: LLM-Driven Verification Layers for Robot Autonomy

    cs.RO 2026-08 conditional novelty 5.0 of 10

    An LLM-driven gate between robot planning and execution labels plans accept, reject, or escalate, reporting 81 percent accuracy and no direct accept/reject errors on small test sets.

  14. SkillCenter: A Large-Scale Source-Grounded Skill Library for Autonomous AI Agents

    cs.AI 2026-07 conditional novelty 5.0 of 10

    SkillCenter provides 216,938 source-grounded skills across 24 domains via an automated LLM-based pipeline, with a downstream evaluation showing skills help agents only when the task exceeds the model's knowledge and r...

  15. LLM Agents Should Employ Security Principles

    cs.CR 2025-05 conditional novelty 5.0 of 10

    A position paper proposing AgentSandbox, a framework that applies Saltzer-Schroeder security principles to LLM agents and reports large attack-success-rate reductions on AgentDojo.

  16. SafeScientist: Toward Risk-Aware Scientific Discoveries by LLM Agents

    cs.AI 2025-05 reject novelty 5.0 of 10

    SafeScientist adds prompt, discussion, tool-use, and output-review safety checks to an AI scientist, with a new domain benchmark, but its reported evaluation is internally inconsistent.

  17. ALRPHFS: Adversarially Learned Risk Patterns with Hierarchical Fast \& Slow Reasoning for Robust Agent Defense

    cs.CR 2025-05 conditional novelty 5.0 of 10

    ALRPHFS builds an adversarially refined library of semantic risk patterns and uses fast retrieval plus slow LLM reasoning to defend LLM agents, reporting best-in-class average accuracy near 80 percent.

  18. Value Compass Benchmarks: A Platform for Fundamental and Validated Evaluation of LLMs Values

    cs.AI 2025-01 conditional novelty 5.0 of 10

    Value Compass Benchmarks is a live, self-evolving platform that scores 33 LLMs across 27 value dimensions from four value systems, aiming to reveal true behavioral alignment with human values.

  19. Semantic Loss Guided Data Efficient Supervised Fine Tuning for Safe Responses in LLMs

    cs.CL 2024-12 conditional novelty 5.0 of 10

    A supervised fine-tuning loss that maximizes an Earth-Mover-Distance-style semantic penalty away from model-generated unsafe responses achieves safety with roughly 100 harmful examples.

  20. Evaluation and Benchmarking of LLM Agents: A Survey

    cs.LG 2025-07 conditional novelty 4.0 of 10

    A review that proposes a two-dimensional taxonomy for evaluating LLM agents and highlights enterprise-specific evaluation gaps.

  21. A Survey on Autonomy-Induced Security Risks in Large Model-Based Agents

    cs.AI 2025-06 conditional novelty 4.0 of 10

    The paper surveys security risks of LLM agents, organizes them into a five-level autonomy taxonomy, and proposes an untested CMDP-based architecture called R2A2.

  22. Forewarned is Forearmed: A Survey on Large Language Model-based Agents in Autonomous Cyberattacks

    cs.NI 2025-05 conditional novelty 4.0 of 10

    A review of LLM-based agents as autonomous cyberattackers, arguing that they lower attack costs, scale up threats, and outpace existing defenses.

  23. Efficient MAP Estimation of LLM Judgment Performance with Prior Transfer

    cs.LG 2025-04 reject novelty 4.0 of 10

    BetaConform estimates LLM ensemble judgment accuracy from few labeled samples via a mixture of Beta-Binomial distributions, conformal-style adaptive stopping, and text-similarity prior transfer, but its theoretical gu...

  24. Agentic Web: Weaving the Next Web with AI Agents

    cs.AI 2025-07 conditional novelty 3.0 of 10

    A position paper defines the Agentic Web as the next web era and proposes a three-dimensional conceptual framework for understanding and building it.

  25. SoK: The Privacy Paradox of Large Language Models: Advancements, Privacy Risks, and Mitigation

    cs.CR 2025-06 conditional novelty 3.0 of 10

    A systematization-of-knowledge survey that categorizes LLM privacy risks into training data, prompts, outputs, and agents, and reviews limitations of current mitigations.

  26. The Scales of Justitia: A Comprehensive Survey on Safety Evaluation of LLMs

    cs.CL 2025-06 conditional novelty 3.0 of 10

    A structured survey of LLM safety evaluation that proposes a why/what/where/how taxonomy and catalogs metrics, datasets, benchmarks, evaluators, and frameworks.

  27. The Science of Evaluating Foundation Models

    cs.CL 2025-02 conditional novelty 3.0 of 10

    A survey-and-checklist proposal that organizes LLM evaluation into an ABCD framework (Algorithm, Big Data, Computation, Domain Expertise) for context-aware, documented assessment.

  28. Large Language Model Safety: A Holistic Survey

    cs.AI 2024-12 conditional novelty 3.0 of 10

    A broad survey of LLM safety that groups the literature into four risk areas and four related areas, with a taxonomy and a public repository of papers, but no new empirical results.

Pith tools