MoHallBench - Visual Similarity Hallucination (Multiple Choice): leaderboard
Metric: Multiple-choice accuracy (%; dual-video four-option choice where the negative clip expects None of these; 32 uniformly sampled frames, temperature 0; visual similarity hallucination). Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 8B Instruct | 59.2 |
| 2 | InternVL3-8B | 58.3 |
| 3 | Qwen 2.5 VL 7B Instruct | 56.6 |
| 4 | Qwen 2 VL 7B Instruct | 52 |
| 5 | Molmo2-8B | 51.2 |
Interactive version: theaggregate.ai/benchmark?slug=mohallbench-visual-similarity-hallucination-multiple-choice · How It Works · Data refreshed daily, snapshot 2026-09-29.