VNU-Bench - Main Claim Comparison: leaderboard
VNU-Bench (Liu et al., University of Florida, 2026) question type in which a multimodal model must compare the main conclusions that different outlets draw about the same news story, given a group of YouTube news videos on one event. Each video is given as 4 selected frames and a cleaned transcript cut to 1,000 characters; zero-shot at temperature 0. Accuracy in percent on 4-option multiple-choice questions; higher is better.
Metric: Accuracy (%; question type T1, 4-option multiple choice). Source: arxiv.org. Saturation forecast: Around March 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 63.24 | #244 |
| 2 | Gemini 2.5 Pro | 61.64 | #262 |
| 3 | Gemini 2.5 Flash | 60.93 | #452 |
| 4 | Qwen 2.5 VL 7B Instruct | 53.68 | #1842 |
| 5 | InternVL3-8B | 52.84 | #1695 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=vnu-bench-main-claim-comparison · How It Works · Data refreshed daily, snapshot 2026-11-12.