Research map / Image generation / Autoregressive and latent generation
LEGION: Learning to Ground and Explain for Synthetic Image Detection
Hengrui Kang, Siwei Wen, Zichen Wen, Junyan Ye, Weijia Li, Peilin Feng, Baichuan Zhou, Bin Wang, Dahua Lin, Linfeng Zhang, Conghui He
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use LEGION MLLM instead of PAL4VST expert model | artifact localization · SynthScars Object | F1: 19.25 -> 29.90 | Improved |
| use LEGION instead of PAL4VST | artifact localization · LOKI (unseen) | mIoU: 52.55 -> 54.52 (Animal) ; 47.34 -> 48.66 (LOKI) | Improved |
| use LEGION instead of LISA-v1-7B | artifact localization · RichHF-18K (unseen) | F1: 21.94 -> 17.41 | Hurt |
| use LEGION instead of LLaVA-v1.6 for explanations | explanation generation · SynthScars | ROUGE-L: 29.61 -> 39.50 | Improved |
| use LEGION instead of LLaVA-v1.6 for explanations | explanation generation · LOKI | ROUGE-L: 16.07 -> 18.55 | Improved |
| use LEGION instead of UniFD for detection | detection (trained on ProGAN) · UniversalFakeDetect GANs | accuracy: 95.25 -> 97.01 | Improved |
| use LEGION instead of FreqNet for detection | detection (trained on ProGAN) · UniversalFakeDetect Deepfakes | accuracy: 97.40 -> 63.37 | Hurt |
| add LEGION-guided prompt revision to regeneration | SD3.5 regeneration · SynthScars test (200 images) | HPS: 31.24 -> 33.36 (6.98%) | Improved |
| add LEGION mask and explanation guidance to inpainting | SDXL inpainting · SynthScars test (200 images) | HPS: 29.57 -> 30.20 (2.14%) | Improved |
| use GPT-4o instead of LEGION for explanations | explanation generation · SynthScars | CSS: 72.60 -> 53.55 | Hurt |
Later papers that compared with this method
1 later accepted paper reports LEGION as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- See and Fix the Flaws: Enabling VLMs and Diffusion Models to Comprehend Visual Artifacts via Agentic Data Synthesis (CVPR 2026 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Spot the Fake: Large Multimodal Model-Based Synthetic Image Detection with Artifact Explanation (NeurIPS 2025 · similarity 0.89)
- CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (CVPR 2025 · similarity 0.88)
- Community Forensics: Using Thousands of Generators to Train Fake Image Detectors (CVPR 2025 · similarity 0.87)
- SALART-VQA: Diagnosing Whether VLMs Understand Salient Artifacts in Generated Images (NeurIPS 2026 · similarity 0.87)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.86)
- A Bias-Free Training Paradigm for More General AI-generated Image Detection (CVPR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
