MoHallBench - Sequential Inference Hallucination Rate (Generative): leaderboard
Metric: Hallucination rate (%; share of free-form descriptions of truncated clips that state or entail the withheld terminal action, judged by a DeepSeek entailment judge with manual review; lower is better). Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Molmo2-8B | 43.3 |
| 2 | Qwen 2.5 VL 7B Instruct | 56.8 |
| 3 | Qwen 3 VL 8B Instruct | 57 |
| 4 | InternVL3-8B | 66.2 |
| 5 | Qwen 2 VL 7B Instruct | 66.8 |
Interactive version: theaggregate.ai/benchmark?slug=mohallbench-sequential-inference-hallucination-rate-generative · How It Works · Data refreshed daily, snapshot 2026-09-29.