VLM-SABRE - Attribute: leaderboard
The attribute subset of VLM-SABRE — items probing whether an attribute judgement tracks the image or a learned prior; accuracy %.
Metric: Accuracy (%). Source: zesearch.github.io. Saturation forecast: Around 2028. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 20 | #79 |
| 2 | Kimi K2.6 | 17 | #104 |
| 3 | Grok 4.3 | 16 | #188 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vlm-sabre-attribute · How It Works · Data refreshed daily, snapshot 2026-10-20.