AEGIS (Academic Image Forensics) - Manipulation Classification: leaderboard
AEGIS (2026) tests whether multimodal models can audit AI-generated or AI-edited images in academic papers, from figure panels across seven scientific image categories and forgeries made with 25 image generators. Macro-F1 for naming the kind of edit made to a highlighted region, in percent; higher is better.
Metric: Macro-F1 (%) over three classes, naming a highlighted edited region, given the original caption, as an insertion, removal or alteration, on AEGIS academic images (seven categories, 39 subtypes; real panels and forgeries from 25 generative models under four forgery strategies), zero-shot minimal prompts, lossless PNG input; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 26 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5.1 | 48.34 | #240 | |
| 2 | Gemini 3 Pro (Preview) | 45.42 | #119 | |
| 3 | Doubao-Seed-1.6 | Thinking | 41.85 | |
| 4 | Llama 4 Maverick | 40.18 | #1110 | |
| 5 | GPT-4.1 | 40.05 | #505 | |
| 6 | Claude Sonnet 4.5 | 39.7 | #261 | |
| 7 | LLaVA-OneVision-72B | 39.42 | #1771 | |
| 8 | Gemini 2.5 Flash | 37.28 | #463 | |
| 9 | Seed-1.6 | 37.08 | #469 | |
| 10 | Qwen 3 VL Plus | 36.89 | #512 | |
| 11 | Qwen-VL-Max (AEGIS checkpoint unspecified) | 36.76 | ||
| 12 | Qwen 2.5 VL 72B Instruct | 35.96 | #841 | |
| 13 | O4 Mini | High | 35.54 | #368 |
| 14 | Gemma 3 27B (IT) | 34.76 | #1324 | |
| 15 | Ministral 3 14B | 33.89 | #1870 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=aegis-academic-image-forensics-manipulation-classification · How It Works · Data refreshed daily, snapshot 2026-11-21.