SaliTrap: leaderboard
Salience bias in everyday commonsense: models have learned to weight explicit stated conditions so heavily that an obvious unstated one gets ignored. The traps are ordinary situations where a human would never need the condition spelled out.
Metric: Overall Trap Avoidance Rate (%). Source: arxiv.org. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 54.8 | #49 |
| 2 | GPT-5.5 | 45.5 | #31 |
| 3 | Claude Opus 4.6 | 44.5 | #63 |
| 4 | GPT-5.4 | 41.2 | #79 |
| 5 | Gemini 2.5 Pro | 27.7 | #152 |
| 6 | GLM-5.1 | 27.1 | #97 |
| 7 | Kimi K2 | 23.9 | #243 |
| 8 | DeepSeek R1 | 23.8 | #251 |
| 9 | DeepSeek V4 Pro | 14.4 | #102 |
| 10 | DeepSeek V4 Flash | 12.1 | #137 |
| 11 | MiniMax-M2.7 | 8.8 | #258 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=salitrap · How It Works · Data refreshed daily, snapshot 2026-10-20.