Research map / Large language models / Language, safety and interpretability
COPYLENS: Towards Copyrighted Characters Infringement Detection via Copyright-Aware Prompt Learning
Yaoyu Jin, Xiaochun Yang, Hong Liu, Leixia Wang, Jian Li, Rui Ding, Bin Wang
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 Token Learning Rate to the meta-prompt (ablation) | CopyLens · COPYCHARS-TEST | Cohen's Kappa: 0.27 -> 0.39 | Improved |
| add Prompt Guidance (ablation) | CopyLens · COPYCHARS-TEST | Cohen's Kappa: 0.39 -> 0.41 | Improved |
| add Masking Mechanism (ablation) | CopyLens · COPYCHARS-TEST | Cohen's Kappa: 0.41 -> 0.53 | Improved |
| add TLR Decay (ablation) | CopyLens · COPYCHARS-TEST | Cohen's Kappa: 0.53 -> 0.58 | Improved |
| add Local Optimum Detection (ablation) | CopyLens · COPYCHARS-TEST | Cohen's Kappa: 0.58 -> 0.62 | Improved |
| optimize detection prompt with CopyLens instead of CopyCat manual prompt | Qwen2-VL-7B detector · COPYCHARS-TEST (avg) | Cohen's Kappa: 0.58 -> 0.62 | Improved |
| use generic LLM prompt optimizers (OPRO, APE, IPO, LLM-OPT) instead of CopyLens | Qwen2-VL-7B detector · COPYCHARS-TEST (avg) | Cohen's Kappa: 0.58 -> 0.19-0.27 | Hurt |
| fine-tune the detector instead of prompt optimization | Qwen2-VL-7B detector · COPYCHARS-TEST (avg) | Cohen's Kappa: 0.58 -> 0.54 | Hurt |
| transfer the Qwen-optimized prompt zero-shot to other detectors | CopyLens prompt · COPYCHARS-TEST, LLaVA-1.5-7B | Cohen's Kappa (%): 45.3 -> 49.7 | Improved |
| re-optimize the prompt for the new detector | CopyLens · COPYCHARS-TEST, LLaVA-1.5-7B | Cohen's Kappa (%): 49.7 -> 53.0 | Improved |
| re-optimize the prompt for GPT-4o-mini-vision instead of transfer | CopyLens · COPYCHARS-TEST, GPT-4o-mini-vision | Cohen's Kappa (%): 57.2 -> 61.4 | Improved |
| CopyLens prompt evaluated on unseen IPs | CopyLens · COPYCHARS-UNSEEN | Cohen's Kappa (%): 55.9 -> 63.3 | 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.
- LLM-OPT — LLMOPT: Learning to Define and Solve General Optimization Problems from Scratch (ICLR 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
- Bridging the Copyright Gap: Do Large Vision-Language Models Recognize and Respect Copyrighted Content? (AAAI 2026 · similarity 0.88)
- Tracking the Copyright of Large Vision-Language Models through Parameter Learning Adversarial Images (ICLR 2025 · similarity 0.86)
- How to Evaluate and Mitigate IP Infringement in Visual Generative AI? (ICML 2025 · similarity 0.85)
- DIS-CO: Discovering Copyrighted Content in VLMs Training Data (ICML 2025 · similarity 0.85)
- Revisiting Tampered Scene Text Detection in the Era of Generative AI (AAAI 2025 · similarity 0.85)
- Echoes of Ownership: Adversarial-Guided Dual Injection for Copyright Protection in MLLMs (CVPR 2026 · 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.
