Research map / Continual learning / Class-incremental
IncreFA: Breaking the Static Wall of Generative Model Attribution
Haotian Qin, Dongliang Chang, Yueying Gao, Yuexuan Tan, Lei Chen, Zhanyu Ma
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 latent replay loss (ablation) | IncreFA baseline · IABench EP1 final task | accuracy (%): 35.24 -> 52.99 | Improved |
| add model-level hierarchical constraint L1 (ablation) | IncreFA · IABench EP1 final task | accuracy (%): 52.99 -> 73.16 | Improved |
| add family-level hierarchical constraint L2 (ablation) | IncreFA · IABench EP1 final task | accuracy (%): 73.16 -> 76.16 | Improved |
| add open-set loss L_u (ablation) | IncreFA · IABench EP1 unseen detection | unseen accuracy (%): 81.94 -> 98.93 | Improved |
| add open-set loss L_u (ablation) | IncreFA · IABench EP1 final task | accuracy (%): 76.16 -> 78.80 | Improved |
| add L1 on top of replay (ablation) | IncreFA · IABench EP1 unseen detection | unseen accuracy (%): 73.98 -> 72.19 | Hurt |
| raise tau from 0.5 to 0.65 (ablation) | IncreFA open-set threshold · IABench EP1 | unseen accuracy (%): 65.01 -> 98.93 | Improved |
| raise tau from 0.65 to 0.90 (ablation) | IncreFA open-set threshold · IABench EP1 | seen accuracy (%): 78.80 -> 41.80 | Hurt |
| increase memory bank from 5 to 150 samples per model (ablation) | IncreFA latent memory bank · IABench EP1 | final accuracy (%): 56.04 -> 77.80 | Improved |
| increase memory bank from 150 to 500 samples per model (ablation) | IncreFA latent memory bank · IABench EP1 | final accuracy (%): 77.80 -> 82.00 | Improved |
| use hierarchical structure with correct family labels instead of wrong D->G family labels (ablation) | IncreFA · IABench EP1 | increment accuracy (%): 71.47 -> 78.80 | Improved |
| replace best baseline with IncreFA under incremental learning | DGR · IABench EP1 final task | accuracy (%): 75.68 -> 78.80 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- Foster — FOSTER: Feature Boosting and Compression for Class-Incremental Learning (ECCV 2022) · better on 1 of 1 shared result
- DualPrompt — DualPrompt: Complementary Prompting for Rehearsal-Free Continual Learning (ECCV 2022) · better on 1 of 1 shared result
- ICaRL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · 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
- GenAI Arena: An Open Evaluation Platform for Generative Models (NeurIPS 2024 · similarity 0.86)
- Learning Visual Generative Priors without Text (CVPR 2025 · similarity 0.85)
- Benchmarking the Attribution Quality of Vision Models (NeurIPS 2024 · similarity 0.85)
- Where's the Liability in the Generative Era? Recovery-based Black-Box Detection of AI-Generated Content (CVPR 2025 · similarity 0.85)
- SDDGR: Stable Diffusion-based Deep Generative Replay for Class Incremental Object Detection (CVPR 2024 · similarity 0.85)
- Revisiting Generative Replay for Class Incremental Object Detection (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.
