KnowHal - Attribute (Positive): leaderboard

Metric: Accuracy (%; attribute-dimension positive questions; 1,800 entity-image samples in 10 domains with paired positive and false-premise negative questions; positive answers scored by standard VQA answer normalization and soft matching, negative answers by a GPT-4o-mini judge (option matching for relation questions); zero-shot). Source: arxiv.org. Saturation forecast: Around December 2026. 14 models tracked.

Top models

#ModelScore
1Seed 2.0 Mini91.8
2Qwen 3 VL 32B Instruct88.9
3Gemini 3 Flash (Preview)84.8
4Kimi K2.584
5GLM-4.6V83.1
6GPT-5.282.3
7Qwen 2.5 VL 72B Instruct81.4
8GPT-5.6 Sol79.4
9InternVL3.5-8B73.4

Interactive version: theaggregate.ai/benchmark?slug=knowhal-attribute-positive · How It Works · Data refreshed daily, snapshot 2026-09-29.