INFACT (Faithfulness) - Clean Accuracy: leaderboard
Metric: Accuracy (0-1, shown times 100) on clean video (Mode I) for INFACT's faithfulness questions that must be answered from the video's visual evidence (static entities and attributes, actions and motions, spatio-temporal relations); zero-shot, default prompts, 16 uniformly sampled frames per video; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 78.4 | #93 |
| 2 | GPT-5.1 | 68.7 | #131 |
| 3 | Qwen 3 VL 32B Instruct | 60.2 | #276 |
| 4 | InternVL3-8B | 57.6 | #606 |
| 5 | Qwen 3 VL 8B Instruct | 55.7 | #401 |
| 6 | Qwen 2.5 VL 32B Instruct | 53.8 | #443 |
| 7 | Qwen 2.5 VL 7B Instruct | 53.1 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=infact-faithfulness-clean-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-11.