KnowHal - Knowledge (Positive): leaderboard

Metric: Accuracy (%; knowledge-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
1Gemini 3 Flash (Preview)79.2
2Seed 2.0 Mini69.2
3Kimi K2.563.4
4GPT-5.6 Sol60.4
5GLM-4.6V56.9
6Qwen 3 VL 32B Instruct56.3
7Qwen 2.5 VL 72B Instruct56.1
8GPT-5.243.3
9InternVL3.5-8B24.5

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