Research map / Image generation / Autoregressive and latent generation
SynerDetect: Hierarchical Synergistic Learning for Generalizable AI-Generated Image Detection
Shuaibo Li, Yijun Yang, Zhaohu Xing, Hongqiu Wang, Pengfei Hao, Xingyu Li, Zekai Liu, Qing Zhang, Lei Zhu
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 Cross-Model Interactive Distillation (generative reconstruction into perceptual encoder) (ablation) | SynerDetect · AIGCDetectBenchmark | mean accuracy (%): 82.52 -> 94.76 | Improved |
| add OT-DCL fusion (ablation) | SynerDetect · AIGCDetectBenchmark | mean accuracy (%): 85.41 -> 94.76 | Improved |
| add prototype-based optimal transport regularizer (ablation) | SynerDetect OT-DCL · AIGCDetectBenchmark | mean accuracy (%): 87.24 -> 94.76 | Improved |
| use dedicated OT-DCL fusion instead of generic fusion or concatenation (ablation) | SynerDetect · AIGCDetectBenchmark | mean accuracy (%): 74.02 (Fusion-B) -> 94.76 | Improved |
| combine perceptual and generative cues versus either alone (ablation) | SynerDetect · AIGCDetectBenchmark | mean accuracy (%): 76.80 / 67.90 / 69.80 -> 94.76 | Improved |
| use simple feature concatenation of CLIP and diffusion features (ablation) | SynerDetect · AIGCDetectBenchmark | mean accuracy (%): 76.80 (CLIP) -> 69.30 | Hurt |
| use learnable Visual-aware Tuner prompts instead of fixed text prompts (ablation) | SynerDetect · AIGCDetectBenchmark | mean accuracy (%): 89.36 -> 94.76 | Improved |
| use both global visual embedding and local CConv features in prompt tuner (ablation) | SynerDetect · AIGCDetectBenchmark | mean accuracy (%): 92.27 / 91.03 -> 94.76 | Improved |
| train SynerDetect on SD v1.4 and test on Chameleon | SynerDetect · Chameleon | accuracy (%): 65.49 -> 71.02 | Improved |
| SynerDetect versus AIDE on GenImage | SynerDetect · GenImage | mean accuracy (%): 86.88 -> 92.39 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CO-SPY — CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (CVPR 2025) · better on 3 of 3 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
- Diversity over Uniformity: Rethinking Representation in Generated Image Detection (CVPR 2026 · similarity 0.90)
- Seeing What Matters: Generalizable AI-generated Video Detection with Forensic-Oriented Augmentation (NeurIPS 2025 · similarity 0.89)
- A Bias-Free Training Paradigm for More General AI-generated Image Detection (CVPR 2025 · similarity 0.88)
- Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated Images (CVPR 2026 · similarity 0.87)
- Community Forensics: Using Thousands of Generators to Train Fake Image Detectors (CVPR 2025 · similarity 0.87)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · 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.
