Research map / Image generation / Autoregressive and latent generation
Dissect and Prune: Enhancing Robustness in AI-Generated Image Detection
Dahye Kim, Jaehyun Choi, Hyun Seok Seong, Seongho Kim, Donghun Lee, Sungwon Yi, Jang-Ho Choi
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 |
|---|---|---|---|
| apply DEAR bilateral pruning plus inpaint-augmented classifier refinement | Corvi · post-processed 9 generators | F.Acc: 47.7 -> 90.2 | Improved |
| apply DEAR bilateral pruning plus inpaint-augmented classifier refinement | Rajan · post-processed 9 generators | F.Acc: 58.2 -> 89.5 | Improved |
| apply DEAR to Corvi | Corvi · post-processed 9 generators | AUC: 79.2 -> 92.3 | Improved |
| apply DEAR to Rajan | Rajan · post-processed 9 generators | AUC: 92.9 -> 97.1 | Improved |
| apply DEAR to Corvi | Corvi · post-processed 9 generators | R.Acc: 97.3 -> 76.4 | Hurt |
| apply DEAR to Rajan | Rajan · post-processed 9 generators | R.Acc: 99.9 -> 95.1 | Hurt |
| apply DEAR to Corvi | Corvi · original 9 generators | F.Acc: 86.5 -> 100.0 (DEAR-c); FLUX F.Acc 21.5 for Corvi | Improved |
| apply DEAR to Rajan | Rajan · original 9 generators | F.Acc: 75.0 -> 99.2 | Improved |
| apply DEAR to Corvi | Corvi · in-the-wild (Chameleon, LOKI, WildRF), post-processed | F.Acc: 27.5 -> 63.5 | Improved |
| apply DEAR to Rajan | Rajan · in-the-wild (Chameleon, LOKI, WildRF), original | F.Acc: 40.0 -> 60.5 | Improved |
| apply DEAR to Corvi | Corvi · in-the-wild (Chameleon, LOKI, WildRF), original | R.Acc: 99.4 -> 89.0 | Hurt |
| use Stay-Positive (non-negative last layer) instead of DEAR | Corvi+ · post-processed 9 generators | F.Acc: 47.7 -> 51.0 (Corvi to Corvi+) vs 90.2 for DEAR-c | Mixed |
Closest papers by meaning
- Dual Data Alignment Makes AI-Generated Image Detector Easier Generalizable (NeurIPS 2025 · similarity 0.86)
- Stay-Positive: A Case for Ignoring Real Image Features in Fake Image Detection (ICML 2025 · similarity 0.86)
- A Bias-Free Training Paradigm for More General AI-generated Image Detection (CVPR 2025 · similarity 0.86)
- Are High-Quality AI-Generated Images More Difficult for Models to Detect? (ICML 2025 · similarity 0.85)
- BIAS-ID: A Framework for Analyzing Transformation Biases in AI-Generated Image Detectors (NeurIPS 2026 · similarity 0.85)
- Orthogonal Subspace Decomposition for Generalizable AI-Generated Image Detection (ICML 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.
