How does information access affect LLM monitors'ability to detect sabotage?

2026-01-28

AI Summary

This study addresses the unclear relationship between the amount of input information and the effectiveness of large language models (LLMs) in detecting alignment risks—such as sabotage behaviors—during monitoring. It reveals a “less-is-more” effect in automated supervision: reducing input information can paradoxically enhance detection performance. To leverage this insight, the authors propose Extract-and-Evaluate (EaE), a hierarchical monitoring framework that decouples information extraction from risk evaluation to optimize the detection pipeline. Evaluated on the BigCodeBench-Sabotage benchmark, EaE improves sabotage detection rates by 16.8 percentage points without increasing false positives. It consistently matches or outperforms baseline methods across diverse settings, and when integrated with information filtering strategies from Gemini and Claude, it achieves Pareto-optimal trade-offs between cost and performance.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Safety and Robustness
• Computer Vision: Large Vision Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Frontier language model agents can exhibit misaligned behaviors, including deception, exploiting reward hacks, and pursuing hidden objectives. To control potentially misaligned agents, we can use LLMs themselves to monitor for misbehavior. In this paper, we study how information access affects LLM monitor performance. While one might expect that monitors perform better when they have access to more of the monitored agents'reasoning and actions, we find that contemporary systems often perform better with less information, a phenomenon we call the less-is-more effect for automated oversight. We demonstrate this phenomenon, and analyze the conditions under which it occurs, in three evaluation environments where agents must conduct sabotage while evading monitors. Leveraging this effect, we introduce extract-and-evaluate (EaE) monitoring -- a new hierarchical approach where one monitor isolates relevant excerpts from the monitored agent's trajectory and another monitor scores these excerpts. In BigCodeBench-Sabotage with a GPT-4.1-mini monitor, EaE improves sabotage detection rates by 16.8 percentage points over the next-best approach. In other settings, EaE either outperforms or is competitive with baselines. In addition, we find that agents unaware of being monitored can be caught much more easily and that monitors scoring longer reasoning traces gain more from information filtering. Lastly, we conduct a cost-performance analysis and find that Gemini and Claude models with monitoring techniques that involve information filtering occupy much of the Pareto frontier.

Problem

• information access
• LLM monitoring
• sabotage detection
• less-is-more effect
• misaligned agents

Innovation

• less-is-more effect
• extract-and-evaluate (EaE)
• LLM monitoring
• sabotage detection
• information filtering