KnowHal - Entity (Positive): leaderboard

Metric: Accuracy (%; entity-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 Mini93
2Gemini 3 Flash (Preview)92.6
3Kimi K2.587.6
4GLM-4.6V81.6
5Qwen 3 VL 32B Instruct79.9
6Qwen 2.5 VL 72B Instruct79.2
7GPT-5.6 Sol70
8GPT-5.259.9
9InternVL3.5-8B38.1

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