Research map / Image generation / Autoregressive and latent generation
UniGenDet: A Unified Generative-Discriminative Framework for Co-Evolutionary Image Generation and Generated Image Detection
Yanran Zhang, Wenzhao Zheng, Yifei Li, Bingyao Yu, Yu Zheng, Lei Chen, Jiwen Lu, Jie Zhou
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 GDUF second-stage detector-guided optimization (ablation) | UniGenDet · FakeClue | accuracy: 40.5 -> 98.0 | Improved |
| add GDUF second-stage optimization (ablation) | UniGenDet · FakeClue | F1: 34.1 -> 97.7 | Improved |
| add SMSA module (ablation) | UniGenDet · FakeClue | accuracy: 95.0 -> 98.0 | Improved |
| add SMSA module (ablation) | UniGenDet · FakeClue | F1: 94.6 -> 97.7 | Improved |
| add SMSA module (ablation) | UniGenDet · FakeClue | ROUGE-L: 50.9 -> 56.3 | Improved |
| add detector-guided generator alignment (stage 2) (ablation) | BAGEL · FID | FID: 22.9 -> 17.5 | Improved |
| add SMSA and full two-stage training on top of BAGEL+GDUF (ablation) | UniGenDet · FID | FID: 19.4 -> 17.5 | Improved |
| replace FakeVLM with UniGenDet under JPEG quality 90 | FakeVLM · FakeClue JPEG 90 | accuracy: 86.8 -> 95.1 | Improved |
| replace FakeVLM with UniGenDet under JPEG quality 50 | FakeVLM · FakeClue JPEG 50 | accuracy: 80.4 -> 91.3 | Improved |
| replace FakeVLM with UniGenDet, no perturbation | FakeVLM · FakeClue | accuracy: 98.6 -> 98.0 | No clear effect |
| replace FakeVLM with UniGenDet on LlamaGen | FakeVLM · ARForensics (LlamaGen) | accuracy: 98.1 -> 89.4 | Hurt |
| replace FakeVLM with UniGenDet on mean | FakeVLM · ARForensics | mean accuracy: 97.1 -> 98.1 | 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.
- D3QE — D3QE: Learning Discrete Distribution Discrepancy-aware Quantization Error for Autoregressive-Generated Image Detection (ICCV 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
- CO-SPY: Combining Semantic and Pixel Features to Detect Synthetic Images by AI (CVPR 2025 · similarity 0.88)
- Generative Unlearning for Any Identity (CVPR 2024 · similarity 0.86)
- Is Artificial Intelligence Generated Image Detection a Solved Problem? (NeurIPS 2025 · similarity 0.85)
- WildFake: A Large-Scale and Hierarchical Dataset for AI-Generated Images Detection (AAAI 2025 · similarity 0.85)
- BIAS-ID: A Framework for Analyzing Transformation Biases in AI-Generated Image Detectors (NeurIPS 2026 · similarity 0.85)
- Towards Universal AI-Generated Image Detection by Variational Information Bottleneck Network (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.
