# Benchmarking Reward Hack Detection in Code Environments via Contrastive Analysis

arXiv: https://arxiv.org/abs/2601.20103 · HTML: https://mlsanity.com/paper/2601.20103/

## Key facts

- Cited by 18 papers in the map (data through 14 Oct 2026).
- Placed in Reward hacking and misspecification, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 40 of its 78 references that are in the map.
- First posted on arXiv 27 Jan 2026.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Reward hacking and misspecification](https://mlsanity.com/alignment/reward-hacking-and-misspecification/)

## Builds on (most cited first, 30 of 40)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code](https://mlsanity.com/paper/2403.07974/) (2024)
- [Show Your Work: Scratchpads for Intermediate Computation with Language Models](https://mlsanity.com/paper/2112.00114/) (2021)
- [DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models](https://mlsanity.com/paper/2512.02556/) (2025)
- [Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2307.15217/) (2023)
- [WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct](https://mlsanity.com/paper/2308.09583/) (2023)
- [Kimi K2: Open Agentic Intelligence](https://mlsanity.com/paper/2507.20534/) (2025)
- [GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models](https://mlsanity.com/paper/2508.06471/) (2025)
- [CodeRL: Mastering Code Generation through Pretrained Models and Deep Reinforcement Learning](https://mlsanity.com/paper/2207.01780/) (2022)
- [The Effects of Reward Misspecification: Mapping and Mitigating Misaligned Models](https://mlsanity.com/paper/2201.03544/) (2022)
- [AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detection](https://mlsanity.com/paper/2310.18961/) (2023)
- [SWE-bench Multimodal: Do AI Systems Generalize to Visual Software Domains?](https://mlsanity.com/paper/2410.03859/) (2024)
- [Crossing the Reward Bridge: Expanding RL with Verifiable Rewards Across Diverse Domains](https://mlsanity.com/paper/2503.23829/) (2025)
- [Natural Emergent Misalignment from Reward Hacking in Production RL](https://mlsanity.com/paper/2511.18397/) (2025)
- [RLTF: Reinforcement Learning from Unit Test Feedback](https://mlsanity.com/paper/2307.04349/) (2023)
- [Enhancing Code LLMs with Reinforcement Learning in Code Generation: A Survey](https://mlsanity.com/paper/2412.20367/) (2024)
- [Feedback Loops With Language Models Drive In-Context Reward Hacking](https://mlsanity.com/paper/2402.06627/) (2024)
- [AA-CLIP: Enhancing Zero-shot Anomaly Detection via Anomaly-Aware CLIP](https://mlsanity.com/paper/2503.06661/) (2025)
- [School of Reward Hacks: Hacking harmless tasks generalizes to misaligned behavior in LLMs](https://mlsanity.com/paper/2508.17511/) (2025)
- [ImpossibleBench: Measuring LLMs' Propensity of Exploiting Test Cases](https://mlsanity.com/paper/2510.20270/) (2025)
- [Co-Evolving LLM Coder and Unit Tester via Reinforcement Learning](https://mlsanity.com/paper/2506.03136/) (2025)
- [Towards Zero-Shot Anomaly Detection and Reasoning with Multimodal Large Language Models](https://mlsanity.com/paper/2502.07601/) (2025)
- [A Comprehensive Survey of Reward Models: Taxonomy, Applications, Challenges, and Future](https://mlsanity.com/paper/2504.12328/) (2025)
- [Writing-Zero: Bridge the Gap Between Non-verifiable Tasks and Verifiable Rewards](https://mlsanity.com/paper/2506.00103/) (2025)
- [Online DPO: Online Direct Preference Optimization with Fast-Slow Chasing](https://mlsanity.com/paper/2406.05534/) (2024)
- [EvilGenie: A Reward Hacking Benchmark](https://mlsanity.com/paper/2511.21654/) (2025)
- [Optimizing Safe and Aligned Language Generation: A Multi-Objective GRPO Approach](https://mlsanity.com/paper/2503.21819/) (2025)

## Built on by (most cited first, 18 of 18)

- [Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges](https://mlsanity.com/paper/2604.13602/) (2026)
- [SpecBench: Measuring Reward Hacking in Long-Horizon Coding Agents](https://mlsanity.com/paper/2605.21384/) (2026)
- [SWE-Marathon: Can Agents Autonomously Complete Ultra-Long-Horizon Software Work?](https://mlsanity.com/paper/2606.07682/) (2026)
- [Detecting Safety Violations Across Many Agent Traces](https://mlsanity.com/paper/2604.11806/) (2026)
- [Do Androids Dream of Breaking the Game? Systematically Auditing AI Agent Benchmarks with BenchJack](https://mlsanity.com/paper/2605.12673/) (2026)
- [Hack-Verifiable Environments: Towards Evaluating Reward Hacking at Scale](https://mlsanity.com/paper/2605.20744/) (2026)
- [Detecting and Suppressing Reward Hacking with Gradient Fingerprints](https://mlsanity.com/paper/2604.16242/) (2026)
- [When the Reward Suite Is Leaky: A Preregistered Causal Contrast of Natural Verifier False Positives in RLVR](https://mlsanity.com/paper/2607.11022/) (2026)
- [Auditing Harness Tampering in Self-Improving Agents](https://mlsanity.com/paper/2609.00069/) (2026)
- [Boiling the Frog: A Multi-Turn Benchmark for Agentic Safety](https://mlsanity.com/paper/2605.22643/) (2026)
- [Likelihood Hacking in Probabilistic Program Synthesis](https://mlsanity.com/paper/2603.24126/) (2026)
- [AgentCompass: A Unified Evaluation Infrastructure for Agent Capabilities](https://mlsanity.com/paper/2607.13705/) (2026)
- [Performance Manipulation: Labor Market Implications in AI-assisted Era](https://mlsanity.com/paper/2604.22230/) (2026)
- [Improving Small Language Models for Code Generation with Reinforcement Learning from Verification Feedback](https://mlsanity.com/paper/2605.30478/) (2026)
- [RH-Detect: A Unified Benchmark for Reward Hacking Detection](https://mlsanity.com/paper/2610.10947/) (2026)
- [CATCH: A Controllable Analysis Testbed for Reward Hacking in Coding RL](https://mlsanity.com/paper/2609.39533/) (2026)
- [SpecGuard: Proving a Task Is Broken Before the Agent Cheats](https://mlsanity.com/paper/2610.09159/) (2026)
- [Checking the Checkers: A Synthesis of Verifier Noise and Robustness in RLVR](https://mlsanity.com/paper/2610.17824/) (2026)

## Abstract (arXiv)

Recent advances in reinforcement learning for code generation have made robust environments essential to prevent reward hacking. As LLMs increasingly serve as evaluators in code-based RL, their ability to detect reward hacking remains understudied. In this paper, we propose a novel taxonomy of reward exploits spanning across 54 categories and introduce TRACE (Testing Reward Anomalies in Code Environments), a synthetically curated and human-verified benchmark containing 517 testing trajectories. Unlike prior work that evaluates reward hack detection in isolated classification scenarios, we contrast these evaluations with a more realistic, contrastive anomaly detection setup on TRACE. Our experiments reveal that models capture reward hacks more effectively in contrastive settings than in isolated classification settings, with GPT-5.2 with highest reasoning mode achieving the best detection rate at 63%, up from 45% in isolated settings on TRACE. Building on this insight, we demonstrate that state-of-the-art models struggle significantly more with semantically contextualized reward hacks compared to syntactically contextualized ones. We further conduct qualitative analyses of model behaviors, as well as ablation studies showing that the ratio of benign to hacked trajectories and analysis cluster sizes substantially impact detection performance. We release the benchmark and evaluation harness to enable the community to expand TRACE and evaluate their models.