Pith. sign in

REVIEW 6 cited by

SHADE-Arena: Evaluating Sabotage and Monitoring in 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 2506.15740 v2 pith:ZKI55P2O submitted 2025-06-17 cs.AI cs.CRcs.LG

classification cs.AIcs.CRcs.LG
keywords sabotageagentsmodelsmonitoringtaskabilityfrontierside
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

As Large Language Models (LLMs) are increasingly deployed as autonomous agents in complex and long horizon settings, it is critical to evaluate their ability to sabotage users by pursuing hidden objectives. We study the ability of frontier LLMs to evade monitoring and achieve harmful hidden goals while completing a wide array of realistic tasks. We evaluate a broad range of frontier LLMs using SHADE (Subtle Harmful Agent Detection & Evaluation)-Arena, the first highly diverse agent evaluation dataset for sabotage and monitoring capabilities of LLM agents. SHADE-Arena consists of complex pairs of benign main tasks and harmful side objectives in complicated environments. Agents are evaluated on their ability to complete the side task without appearing suspicious to an LLM monitor. When measuring agent ability to (a) complete the main task, (b) complete the side task, and (c) avoid detection, we find that the best performing frontier models score 27% (Claude 3.7 Sonnet) and 15% (Gemini 2.5 Pro) as sabotage agents when overseen by Claude 3.6 Sonnet. For current frontier models, success on the side task relies heavily on having access to a hidden scratchpad that is not visible to the monitor. We also use SHADE-Arena to measure models' monitoring abilities, with the top monitor (Gemini 2.5 Pro) achieving an AUC of 0.87 at distinguishing benign and malign transcripts. We find that for now, models still struggle at sabotage due to failures in long-context main task execution. However, our measurements already demonstrate the difficulty of monitoring for subtle sabotage attempts, which we expect to only increase in the face of more complex and longer-horizon tasks.

Discussion (0). Sign in to comment.

Forward citations

Cited by 6 Pith papers

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

  1. Chain-of-Thought Monitoring Can Be Unreliable in Implicit-Influence Settings

    cs.AI 2026-08 conditional novelty 7.0 of 10

    A matched benchmark across four task formats and seven reasoning models shows chain-of-thought monitors detect implicit behavioral influences far worse than explicit ones, with detection dropping to as low as 5 percen...

  2. ResearchArena: Evaluating Sabotage and Monitoring in Automated AI R&D

    cs.AI 2026-07 conditional novelty 7.0 of 10

    In a four-task benchmark of AI research agents, sabotage planted in fine-tuning data is detected less than half the time even by monitors that can run the submitted artifact.

  3. Agent-Safety Evaluations as Load-Bearing Evidence: A Vendor-Neutral, Cross-Harness Reconstructability Metric

    cs.SE 2026-07 conditional novelty 6.0 of 10

    A property-level reconstructability metric and Evidence Sufficiency Card show that traces sharing a surface reading can differ sharply in evidence sufficiency, and that replay preconditions often fail.

  4. 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.

  5. Reliable Weak-to-Strong Monitoring of LLM Agents

    cs.AI 2025-08 conditional novelty 6.0 of 10

    Monitor scaffolding, not monitor awareness or omniscience, drives detection reliability, and a hybrid chunked monitor lets weak models supervise strong LLM agents.

  6. Adversarial Prompting Framework for AI Safety Assessment

    cs.CR 2026-07 reject novelty 4.0 of 10

    An adversarial prompt testing framework with a five-level attack taxonomy and a composite harmfulness score is proposed; the paper claims encoded prompts bypass safety filters most often.

Pith tools