Research map / Image generation / Autoregressive and latent generation
FakeXplain: AI-Generated Image Detection via Human-Aligned Grounded Reasoning
Yikun Ji, Yan Hong, Qi Fan, jun lan, Huijia Zhu, Weiqiang Wang, Liqing Zhang, Jianfu Zhang
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 |
|---|---|---|---|
| add GRPO stage after SFT (ablation) | FakeXplainer 32B · FakeXplained | accuracy: 0.893 -> 0.982 | Improved |
| add SFT fine-tuning to base MLLM (ablation) | Qwen2.5-VL-32B · FakeXplained | accuracy: 0.734 -> 0.893 | Improved |
| use progressive localization weight instead of constant wG=1 (ablation) | FakeXplainer 32B · FakeXplained | IoU: 0.265 -> 0.360 | Improved |
| use progressive localization weight instead of constant wG=1 (ablation) | FakeXplainer 32B · FakeXplained | accuracy: 0.937 -> 0.982 | Improved |
| use progressive localization weight instead of constant wG=0.5 (ablation) | FakeXplainer 32B · FakeXplained | IoU: 0.223 -> 0.360 | Improved |
| add bounding box annotations to training data (ablation) | FakeXplainer 32B · FakeXplained | accuracy: 0.952 -> 0.982 | Improved |
| add captions to training data (ablation) | FakeXplainer 32B · FakeXplained | IoU: 0.265 -> 0.360 | Improved |
| add captions to training data (ablation) | FakeXplainer 32B · FakeXplained | accuracy: 0.942 -> 0.982 | Improved |
| add image tags to training data (ablation) | FakeXplainer 32B · FakeXplained | IoU: 0.358 -> 0.360 | No clear effect |
| add region-level annotations beyond binary labels (ablation) | FakeXplainer 32B · FakeXplained | accuracy: 0.937 -> 0.982 | Improved |
| scale model from 3B to 7B (ablation) | FakeXplainer · FakeXplained | accuracy: 0.842 -> 0.958 | Improved |
| scale model from 7B to 32B (ablation) | FakeXplainer · FakeXplained | accuracy: 0.958 -> 0.982 | Improved |
Closest papers by meaning
- Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated Images (CVPR 2026 · similarity 0.88)
- Spot the Fake: Large Multimodal Model-Based Synthetic Image Detection with Artifact Explanation (NeurIPS 2025 · similarity 0.88)
- Community Forensics: Using Thousands of Generators to Train Fake Image Detectors (CVPR 2025 · similarity 0.88)
- Stay-Positive: A Case for Ignoring Real Image Features in Fake Image Detection (ICML 2025 · similarity 0.87)
- WildFake: A Large-Scale and Hierarchical Dataset for AI-Generated Images Detection (AAAI 2025 · similarity 0.87)
- CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (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.
