EVID-Bench - Narrative Fabrication: leaderboard

Metric: Point-level accuracy (%) on the 27 narrative fabrication videos (multi-source editing: footage spliced from several sources into a false story), 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 December 2026. 9 models tracked.

Top models

#ModelScore
1GPT-5.582.24
2Claude Opus 4.676.64
3GPT-5.473.83
4Qwen 3.5 Plus73.8
5Claude Sonnet 4.672.9
6Gemini 3 Flash72.9
7Gemini 3.1 Pro (Preview)71
8Qwen 3 VL 235B A22B Instruct71
9GPT-5.4 Mini64.49

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