EVID-Bench - Manipulative Montage: leaderboard

Metric: Point-level accuracy (%) on the 25 manipulative montage videos (single-source editing: a misleading montage of one source), 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 February 2027. 9 models tracked.

Top models

#ModelScore
1GPT-5.574
2Claude Opus 4.670
3Gemini 3 Flash70
4Claude Sonnet 4.667
5Gemini 3.1 Pro (Preview)66
6GPT-5.466
7Qwen 3.5 Plus64
8Qwen 3 VL 235B A22B Instruct60
9GPT-5.4 Mini58

Interactive version: theaggregate.ai/benchmark?slug=evid-bench-manipulative-montage · How It Works · Data refreshed daily, snapshot 2026-09-29.