Reward Hacking Benchmark (RHB) — leaderboard
Agentic reward-hacking benchmark with multi-step tool-use tasks containing naturalistic shortcut opportunities such as skipping verification, exploiting metadata, or tampering with evaluation-relevant functions.
Metric: Exploit rate (%) (self-reported). Source: benchmarklist.com. Status: saturation imminent. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O3 | 11.8 |
| 2 | O4 Mini | 8.4 |
| 3 | O3 Mini | 7.1 |
| 4 | O1 | 6.8 |
| 5 | Gemini 2.5 Pro (Preview 06-05) | 4.6 |
| 6 | Claude 3.7 Sonnet | 3.9 |
| 7 | GPT-4o | 0.9 |
| 8 | Gemini 2.5 Flash (Preview 09-2025) | 0.8 |
| 9 | Claude 3.5 Sonnet | 0.6 |
| 10 | DeepSeek V3 | 0.6 |
| 11 | Claude Sonnet 4.5 | 0 |
| 12 | Claude Opus 4.5 | 0 |
Interactive version: theaggregate.ai/benchmark?slug=reward-hacking-benchmark-rhb · How the rankings work · Data refreshed daily, snapshot 2026-07-22.