HINTBench - Coarse Risk-Step Localization (Strict): leaderboard
HINTBench (2026) asks an LLM to audit long-horizon agent trajectories for intrinsic, non-attack risks such as acting beyond what the user authorized or skipping confirmations before irreversible actions. This board scores locating the risky steps and naming their constraint category correctly, as strict F1 in percent, higher is better.
Metric: Strict F1 (%) for coarse risk-step localization: a predicted risky step matches a gold step within plus or minus 3 steps and the predicted primary constraint category must match the gold one, over the 536 synthetic HINTBench agent trajectories (400 risky, 136 safe, 24 steps on average, unanimously accepted by three human verifiers) audited zero-shot by a general LLM for intrinsic, non-attack risk under a five-constraint taxonomy (goal, capability, factual, procedural, state constraints), each model at its stated reasoning setting; higher is better. Source: arxiv.org. Saturation forecast: Around September 2027. 13 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 | Thinking, 32K | 46.32 | #173 |
| 2 | GPT-5.5 | High | 45.96 | #54 |
| 3 | Claude Opus 4.6 | Thinking, 4K | 45.23 | #113 |
| 4 | Kimi K2.5 | 43.63 | #266 | |
| 5 | GLM-5 | 41.19 | #269 | |
| 6 | ERNIE 5.0 | 40.67 | #479 | |
| 7 | DeepSeek R1 | 40.25 | #482 | |
| 8 | MiniMax-M2.5 | 39.88 | #600 | |
| 9 | Qwen 3 235B A22B | 36.28 | #628 | |
| 10 | DeepSeek V3.2 | 31.93 | #419 | |
| 11 | Qwen 3 32B | 23.2 | #972 | |
| 12 | Qwen 3 14B | 22.87 | #1269 | |
| 13 | Qwen 3 8B | 14.34 | #1873 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=hintbench-coarse-risk-step-localization-strict · How It Works · Data refreshed daily, snapshot 2026-11-21.