IslamicLegalBench (Zero-Shot) - False Premise Acceptance: leaderboard
Metric: Acceptance rate (%) on the 51 false-premise queries (a misattributed school or author, a wrong period or fabricated conditions planted in questions of tasks T1 to T3 and T7): share the model answers as if the premise were true instead of challenging it, judged by an o3 judge, temperature 0, zero-shot prompting; lower is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 5.88 | #138 |
| 2 | Gemini 2.5 Pro | 9.8 | #145 |
| 3 | Grok 4 | 11.76 | #169 |
| 4 | GPT-5 | 17.65 | #91 |
| 5 | Qwen 3 235B A22B 2507 Instruct | 39.22 | #291 |
| 6 | Llama 4 Maverick Instruct FP8 | 45.1 | #403 |
| 7 | DeepSeek R1 | 50.98 | #245 |
| 8 | Llama 3.1 8B Instruct | 74.51 | #1018 |
| 9 | GPT-OSS-120B | 80.39 | #330 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=islamiclegalbench-zero-shot-false-premise-acceptance · How It Works · Data refreshed daily, snapshot 2026-10-11.