# From Reward-Hack Activations to Agentic Risk States: Context-Calibrated Mechanistic Monitoring in LLM Agents

arXiv: https://arxiv.org/abs/2606.06223 · HTML: https://mlsanity.com/paper/2606.06223/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 12 of its 16 references that are in the map.
- First posted on arXiv 4 Jun 2026.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)

## Builds on (most cited first, 12 of 12)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents](https://mlsanity.com/paper/2207.01206/) (2022)
- [ALFWorld: Aligning Text and Embodied Environments for Interactive Learning](https://mlsanity.com/paper/2010.03768/) (2020)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [Reasoning Models Don't Always Say What They Think](https://mlsanity.com/paper/2505.05410/) (2025)
- [Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation](https://mlsanity.com/paper/2503.11926/) (2025)
- [Convergent Linear Representations of Emergent Misalignment](https://mlsanity.com/paper/2506.11618/) (2025)
- [School of Reward Hacks: Hacking harmless tasks generalizes to misaligned behavior in LLMs](https://mlsanity.com/paper/2508.17511/) (2025)
- [Monitoring Emergent Reward Hacking During Generation via Internal Activations](https://mlsanity.com/paper/2603.04069/) (2026)
- [SafePro: Evaluating the Safety of Professional-Level AI Agents](https://mlsanity.com/paper/2601.06663/) (2026)
- [Noticing the Watcher: LLM Agents Can Infer CoT Monitoring from Blocking Feedback](https://mlsanity.com/paper/2603.16928/) (2026)
- [Guideline-Grounded Evidence Accumulation for High-Stakes Agent Verification](https://mlsanity.com/paper/2603.02798/) (2026)

## Built on by (most cited first, 3 of 3)

- [Monitoring and Discovering Reward Hacking with Internal Representations during LLM Evaluations](https://mlsanity.com/paper/2609.19101/) (2026)
- [Measuring Harness-Induced Belief Divergence in Multi-Step LLM Agents](https://mlsanity.com/paper/2607.04528/) (2026)
- [From Probe Scores to Alarm Policies: Operational Validity of Activation Monitors for Language-Model Agents](https://mlsanity.com/paper/2610.04575/) (2026)

## Abstract (arXiv)

Language-model agents act through repeated cycles of observation, reasoning, and action selection, making safety monitoring depend on both internal model state and environment context. We study reward-hacking monitors in ReAct-style agents acting in Gameable ALFWorld and WebShop. Agents are instrumented with activation-based reward-hack scores, token-level entropy, and decision-context features. We find that adapters fine-tuned on textitSchool-of-Reward-Hacks dataset can transfer reward-hack tendencies into agentic action selection, especially when the environment exposes proxy-reward affordances. However, mitigating such behavior cannot rely on activation dynamics alone. High reward-hack activation identifies a latent policy state, but does not necessarily imply an immediate exploit action. Across next-step prediction tasks, entropy and context-calibrated internal features improve risk estimation over reward-hack activation alone. Activation-direction steering further reduces proxy-exploit behavior in selected mixed-adapter regimes. Overall, our results support context-calibrated internal monitoring for agents: reward-hack activation identifies a latent policy state, while entropy and decision context help determine when that state becomes risky action.