EVID-Bench - Selective Omission: leaderboard
Metric: Point-level accuracy (%) on the 25 selective omission videos (single-source editing: key context cut out), using the retrieval-augmented verification pipeline of the paper (chain-of-thought analysis, up to six YouTube search-verify-reflect rounds, 64 frames at 480p, temperature 0); each video has 3 to 5 annotated misinformation points, matched by a majority of three LLM judges; mean of three runs; higher is better. Source: arxiv.org. Saturation forecast: Around March 2027. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 72.92 |
| 2 | Claude Opus 4.6 | 69.79 |
| 3 | Gemini 3.1 Pro (Preview) | 67.7 |
| 4 | Claude Sonnet 4.6 | 66.67 |
| 5 | Gemini 3 Flash | 65.6 |
| 6 | GPT-5.4 | 64.58 |
| 7 | Qwen 3 VL 235B A22B Instruct | 61.5 |
| 8 | Qwen 3.5 Plus | 58.3 |
| 9 | GPT-5.4 Mini | 57.29 |
Interactive version: theaggregate.ai/benchmark?slug=evid-bench-selective-omission · How It Works · Data refreshed daily, snapshot 2026-09-29.