Research map / Vision-language and multimodal / CLIP and zero-shot
Rethinking Open-world Prompt Tuning: A Systematic Framework for Evaluation and Optimization
Mengwei Li, Zilei Wang, Yixin 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 |
|---|---|---|---|
| replace trained GMoP detector with training-free ERF | SkipTuning · 11 datasets avg | AUROC: 90.92 -> 95.26 | Improved |
| use ERF instead of NegLabel | SkipTuning · 11 datasets avg | AUROC: 86.06 -> 95.26 | Improved |
| replace GMoP detector with ERF | SkipTuning · 11 datasets avg | Opt-h: 64.12 -> 65.75 | Improved |
| replace GMoP detector with ERF | TCP · 11 datasets avg | Opt-h: 61.84 -> 63.07 | Improved |
| replace GMoP detector with ERF | GMoP · 11 datasets avg | Opt-h: 60.38 -> 60.45 | No clear effect |
| add GDM (gated dual-merging) on top of ERF | SkipTuning · 11 datasets avg | Opt-h: 65.75 -> 67.16 | Improved |
| add GDM on top of ERF | GMoP · 11 datasets avg | Opt-h: 60.45 -> 63.34 | Improved |
| add GDM to prompt tuning method | SkipTuning · 11 datasets avg base-to-new | H: 80.89 -> 82.04 | Improved |
| add GDM to prompt tuning method | TCP · 11 datasets avg base-to-new | H: 79.19 -> 80.93 | Improved |
| add GDM to GMoP | GMoP · 11 datasets avg base-to-new | H: 78.07 -> 80.04 | Improved |
| gate weight merging by delta-D instead of merging WiSE-FT on all datasets (ablation) | SkipTuning · 11 datasets avg | accuracy: +0.59% | Improved |
| add logits merging to gated weight merging (ablation) | SkipTuning · 11 datasets avg | accuracy: +0.58% | 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.
- SkipTuning — Skip Tuning: Pre-trained Vision-Language Models are Effective and Efficient Adapters Themselves (CVPR 2025) · better on 3 of 3 shared results
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
- OpenworldAUC: Towards Unified Evaluation and Optimization for Open-world Prompt Tuning (ICML 2025 · similarity 0.95)
- DeCoOp: Robust Prompt Tuning with Out-of-Distribution Detection (ICML 2024 · similarity 0.92)
- Learning to Learn Better Visual Prompts (AAAI 2024 · similarity 0.91)
- O-TPT: Orthogonality Constraints for Calibrating Test-time Prompt Tuning in Vision-Language Models (CVPR 2025 · similarity 0.90)
- Weak Distribution Detectors Lead to Stronger Generalizability of Vision-Language Prompt Tuning (AAAI 2024 · similarity 0.90)
- DPC: Dual-Prompt Collaboration for Tuning Vision-Language Models (CVPR 2025 · similarity 0.90)
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.
