MoHallBench - Action-Action Hallucination (Multiple Choice): leaderboard
Metric: Multiple-choice accuracy (%; four-option multiple choice (ground truth, hard hallucination distractor, random distractor, None of these); 32 uniformly sampled frames, temperature 0; action-action transition hallucination). Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL3-8B | 72.6 |
| 2 | Qwen 2 VL 7B Instruct | 72.3 |
| 3 | Molmo2-8B | 72.1 |
| 4 | Qwen 3 VL 8B Instruct | 70 |
| 5 | Qwen 2.5 VL 7B Instruct | 65.2 |
Interactive version: theaggregate.ai/benchmark?slug=mohallbench-action-action-hallucination-multiple-choice · How It Works · Data refreshed daily, snapshot 2026-09-29.