Research map / Large language models / AI and society
From Imitation to Innovation: The Emergence of AI's Unique Artistic Styles and the Challenge of Copyright Protection
Zexi Jia, Chuanwei Huang, Yeshuang Zhu, Hongyan Fei, Ying Deng, Zhiqiang Yuan, Jiapei Zhang, Jinchao Zhang, 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 description-guided cross-modal distillation on top of CLIP features | CLIP k-means · Artist style clustering, WikiART | ACC: 36.09 -> 41.45 | Improved |
| raise alpha (cluster entropy weight) from 2 to 3 (ablation) | DGC · Artist style clustering, WikiART | ACC: 35.38 -> 41.45 | Improved |
| raise alpha from 3 to 5 (ablation) | DGC · Artist style clustering, WikiART | ACC: 41.45 -> 40.29 | Hurt |
| raise neighbor count N from 5 to 10 (ablation) | DGC · Artist style clustering, WikiART | ACC: 39.78 -> 41.45 | Improved |
| raise neighbor count N from 10 to 15 (ablation) | DGC · Artist style clustering, WikiART | ACC: 41.45 -> 40.73 | Hurt |
| use CLIP instead of BLIP-2 features (ablation) | DGC · Artist style clustering, WikiART | ACC: 1.78% drop | Improved |
| use CLIP instead of SigLip features (ablation) | DGC · Artist style clustering, WikiART | ACC: 2.54% decrease | Improved |
| add MLLM chain-of-thought reasoning (ArtBulb) over DGC clustering results | DGC · AICD copyright judgment (average) | accuracy: 0.81 -> 0.86 | Improved |
| add MLLM chain-of-thought reasoning (ArtBulb) over DGC clustering results | DGC · AICD copyright judgment (average) | F1: 0.79 -> 0.88 | Improved |
| guide the MLLM with DGC clustering results | GPT-4o · AICD explanation quality (AHR) | expert rating: 2.05 -> 3.70 | Improved |
| guide the MLLM with DGC clustering results | Qwen2-VL · AICD explanation quality (AHR) | expert rating: 1.80 -> 3.55 | Improved |
| use GPT-4o instead of Qwen2-VL or DeepSeek-VL as the MLLM (ablation) | ArtBulb · AICD test set | average accuracy: 0.86 vs 0.81 and 0.74 | Improved |
Closest papers by meaning
- Rethinking Artistic Copyright Infringements In the Era Of Text-to-Image Generative Models (ICLR 2025 · similarity 0.90)
- Stylebreeder: Exploring and Democratizing Artistic Styles through Text-to-Image Models (NeurIPS 2024 · similarity 0.84)
- Adversarial Perturbations Cannot Reliably Protect Artists From Generative AI (ICLR 2025 · similarity 0.84)
- Bridging the Copyright Gap: Do Large Vision-Language Models Recognize and Respect Copyrighted Content? (AAAI 2026 · similarity 0.84)
- How to Evaluate and Mitigate IP Infringement in Visual Generative AI? (ICML 2025 · similarity 0.82)
- Social Reward: Evaluating and Enhancing Generative AI through Million-User Feedback from an Online Creative Community (ICLR 2024 · similarity 0.82)
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.
