Research map / Image generation / Autoregressive and latent generation
GenShield: Unified Detection and Artifact Correction for AI-Generated Images
Zhipei Xu, Xuanyu Zhang, Youmin Xu, Qing Huang, Shen Chen, Taiping Yao, Shouhong Ding, Jian 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 |
|---|---|---|---|
| jointly train artifact correction with detection (vs detection only) (ablation) | GenShield · Holmes-Set | accuracy (%): 96.4 -> 98.8 | Improved |
| jointly train detection with correction (vs correction only) (ablation) | GenShield · SynthScars | HPSv3: 5.39 -> 6.20 | Improved |
| add multi-step Visual CoT self-correction training (ablation) | GenShield · SynthScars | HPSv3: 5.99 -> 6.20 | Improved |
| add VCoT training on detection (ablation) | GenShield · Holmes-Set | accuracy (%): 98.6 -> 98.8 | Improved |
| use two-stage curriculum instead of Stage 1 only (ablation) | GenShield · SynthScars | HPSv3: 5.93 -> 6.20 | Improved |
| use two-stage curriculum instead of Stage 2 only (ablation) | GenShield · Holmes-Set | accuracy (%): 97.6 -> 98.8 | Improved |
| use multi-step correction instead of single-step (ablation) | GenShield · SynthScars GPT-assisted evaluation | mean artifact score: 0.23 -> 0.15 | Improved |
| use multi-step correction instead of single-step (ablation) | GenShield · SynthScars HPSv3 | HPSv3: 6.23 -> 6.20 | No clear effect |
| fine-tune BAGEL with GenShield joint training | BAGEL · SynthScars GPT-assisted evaluation | mean artifact score: 0.74 -> 0.15 | Improved |
| apply JPEG compression QF=70 to test images | GenShield vs baselines · Holmes-Set | accuracy (%): 98.8 -> 97.8 | No clear effect |
Closest papers by meaning
- Is Artificial Intelligence Generated Image Detection a Solved Problem? (NeurIPS 2025 · similarity 0.88)
- A Bias-Free Training Paradigm for More General AI-generated Image Detection (CVPR 2025 · similarity 0.87)
- Stay-Positive: A Case for Ignoring Real Image Features in Fake Image Detection (ICML 2025 · similarity 0.86)
- BIAS-ID: A Framework for Analyzing Transformation Biases in AI-Generated Image Detectors (NeurIPS 2026 · similarity 0.85)
- Spot the Fake: Large Multimodal Model-Based Synthetic Image Detection with Artifact Explanation (NeurIPS 2025 · similarity 0.85)
- CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (CVPR 2025 · similarity 0.85)
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.
