EVID-Bench - Identity Swap: leaderboard

Metric: Point-level accuracy (%) on the 24 identity swap videos (AI generation: the identity of a person replaced by a generated one), 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 2028. 9 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)30.6
2Qwen 3.5 Plus22.2
3Gemini 3 Flash13.9
4GPT-5.511.11
5Claude Opus 4.68.33
6Qwen 3 VL 235B A22B Instruct6.9
7Claude Sonnet 4.65.56
8GPT-5.4 Mini1.39
9GPT-5.40

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