KnowHal - Entity (Negative): leaderboard

Metric: Accuracy (%; entity-dimension negative (false-premise) 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 2034. 14 models tracked.

Top models

#ModelScore
1Qwen 3 VL 32B Instruct52.9
2Qwen 2.5 VL 72B Instruct45.1
3Gemini 3 Flash (Preview)40.1
4Kimi K2.531.8
5InternVL3.5-8B31.7
6GPT-5.228.7
7Seed 2.0 Mini27.1
8GLM-4.6V24.7
9GPT-5.6 Sol22.3

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