Research map / Image generation / Autoregressive and latent generation
Diversity over Uniformity: Rethinking Representation in Generated Image Detection
Qinghui He, Haifeng Zhang, Qiao Qin, Bo Liu, Xiuli Bi, Bin Xiao
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 Cue Information Bottleneck (with weight regularizer) (ablation) | DoU · average over all test sets | ACC: 87.81% -> 89.72% | Improved |
| add Cue Information Bottleneck (with weight regularizer) (ablation) | DoU · average over all test sets | AP: 87.32% -> 99.38% | Improved |
| add Cue Information Bottleneck (ablation) | DoU · cross-generator | ACC: 85.56% -> 85.99% | No clear effect |
| add anti-feature-collapse learning (HSIC decorrelation) (ablation) | DoU · cross-generator | ACC: 85.56% -> 91.15% | Improved |
| combine CIB and AFCL (ablation) | DoU · average over all test sets | ACC: 87.81% -> 92.81% | Improved |
| add CIB on top of AFCL (ablation) | DoU · cross-generator | ACC: 91.15% -> 90.02% | Hurt |
| add CIB on top of AFCL (ablation) | DoU · average over all test sets | ACC: 91.60% -> 92.81% | Improved |
| replace VIB-Net with DoU | generated image detector · average over all test sets | ACC: +5.68% ACC, +3.39% AP | Improved |
| replace CLIPping baseline with DoU | generated image detector · average over all test sets | AP: +12.20% AP, +5.00% ACC | Improved |
| reduce training data from full set to 0.1% (320 samples) (ablation) | DoU · few-shot | ACC: 92.81% -> 80.98% | Hurt |
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 1 of 1 shared result
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
- Community Forensics: Using Thousands of Generators to Train Fake Image Detectors (CVPR 2025 · similarity 0.90)
- Seeing What Matters: Generalizable AI-generated Video Detection with Forensic-Oriented Augmentation (NeurIPS 2025 · similarity 0.90)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.89)
- A Bias-Free Training Paradigm for More General AI-generated Image Detection (CVPR 2025 · similarity 0.88)
- BIAS-ID: A Framework for Analyzing Transformation Biases in AI-Generated Image Detectors (NeurIPS 2026 · similarity 0.88)
- BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation (NeurIPS 2026 · similarity 0.87)
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.
