SABRE-Prior - Texture: leaderboard
The Texture subset of SABRE-Prior: a familiar object before and after its canonical material is edited into a counterfactual one, probed with four yes-or-no questions; a case counts only when all four answers are right. Percent of 100 cases; higher is better.
Metric: Strict All4 accuracy (%; a case counts only when all four yes/no probes on the base and edited image are right: a familiar object given a counterfactual material; zero-shot, greedy decoding where supported, 8-16 output tokens for API models; 100 generated or edited cases per subset that survived filtering against Gemini 3.5 Flash and human review). Source: arxiv.org. Saturation forecast: Around August 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.6 (Non-reasoning) | 52 | #178 (Kimi K2.6) |
| 2 | Qwen 3.5 27B (Non-reasoning) | 46 | #359 (Qwen 3.5 27B) |
| 3 | Claude Sonnet 4.6 | 40 | #161 |
| 4 | Grok 4.3 | 28 | #333 |
| 5 | GPT-5.4 (Non-reasoning) | 28 | #123 (GPT-5.4) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sabre-prior-texture · How It Works · Data refreshed daily, snapshot 2026-11-13.