Research map / Vision-language and multimodal / CLIP and zero-shot
Authorize-on-Demand: Dynamic Authorization with Legality-Aware Intellectual Property Protection for VLMs
Lianyu Wang, Meng Wang, Huazhu Fu, Daoqiang Zhang
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 |
|---|---|---|---|
| use AoD-IP instead of IP-CLIP | CLIP IP protection · Office-31 | W(u-a): 74.84 -> 79.27 | Improved |
| use AoD-IP instead of IP-CLIP | CLIP IP protection · Office-Home-65 | W(u-a): 55.10 -> 63.47 | Improved |
| use AoD-IP instead of IP-CLIP | CLIP IP protection · Mini-DomainNet | W(u-a): 54.68 -> 58.49 | Improved |
| use AoD-IP instead of HNTL | CLIP IP protection · Office-Home-65 | Drop on authorized domain: 19.38 -> 0.13 | Improved |
| use AoD-IP instead of IP-CLIP (authorization application) | CLIP IP protection · Office-31 | D(u-a): 58.78 -> 69.39 | Improved |
| use AoD-IP instead of IP-CLIP (authorization application) | CLIP IP protection · Mini-DomainNet | D(u-a): 48.39 -> 56.59 | Improved |
| add AoD-IP protection vs unprotected SL-CLIP | CLIP · Office-Home-65 | Drop on unauthorized domain: 0 -> 74.57 drop | Improved |
| add AoD-IP protection vs unprotected SL-CLIP | CLIP · Office-Home-65 | Drop on authorized domain: 0.13 | No clear effect |
| IP-CLIP vs AoD-IP on authorized-domain drop | IP-CLIP · Office-Home-65 | Drop on authorized domain: 0.20 vs 0.13 | No clear effect |
| use AoD-IP instead of HNTL (authorization application) | CLIP IP protection · Mini-DomainNet | D(u-a): 1.39 -> 56.59 | Improved |
| switch to extended domain with new credential token (no retraining) | AoD-IP · Office-31 | legality accuracy: Re 96.3-98.8 across tasks | Improved |
| use AoD-IP vs best prior in Office-31 Drop on authorized domain | IP-CLIP · Office-31 | Drop on authorized domain: 0.00 vs 0.00 | No clear effect |
Closest papers by meaning
- Vision-Language Model IP Protection via Prompt-based Learning (CVPR 2025 · similarity 0.85)
- Palette: A Modular, Controllable, and Efficient Framework for On-demand Authorized Safety Alignment Relaxation in LLMs (NeurIPS 2026 · similarity 0.81)
- MAP: MAsk-Pruning for Source-Free Model Intellectual Property Protection (CVPR 2024 · similarity 0.80)
- DRIFT: Dynamic Rule-Based Defense with Injection Isolation for Securing LLM Agents (NeurIPS 2025 · similarity 0.80)
- Permissioned LLMs: Enforcing Access Control in Large Language Models (NeurIPS 2025 · similarity 0.80)
- GuardReasoner-VL: Safeguarding VLMs via Reinforced Reasoning (NeurIPS 2025 · similarity 0.80)
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.
