Research map / Vision-language and multimodal / CLIP and zero-shot
Neutral-Reference Prompting for Vision–Language Models
Senmao Tian, Xiang Wei, Shunli 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 |
|---|---|---|---|
| add NeRP correction | MMRL · Base-to-novel average | Novel accuracy: 77.16 -> 78.04 | Improved |
| add NeRP correction | CoCoOp · Base-to-novel average | Novel accuracy: 71.69 -> 72.99 | Improved |
| add NeRP correction | MMRL · Base-to-novel EuroSAT | Novel accuracy: 80.17 -> 85.17 | Improved |
| add NeRP correction | MMRL · Cross-dataset average | Accuracy: 67.25 -> 68.00 | Improved |
| add NeRP correction | MMA · Domain generalization average | Accuracy: 60.48 -> 61.95 | Improved |
| add NeRP correction | CoCoOp · Cross-dataset FGVCAircraft | Accuracy: 22.94 -> 23.70 | Improved |
| add NeRP correction | PromptSRC · Cross-dataset average | Accuracy: 65.81 -> 66.91 | Improved |
| use finer calibration search step (1e-3 vs 1e-2) (ablation) | MMRL+NeRP · 10-dataset average, E=2 | Accuracy: 76.80 -> 76.88 | Improved |
| use finer calibration search step (1e-3 vs 1e-2) (ablation) | MMRL+NeRP · 10-dataset average, E=5 | Flip error rate: 25.44 -> 19.91 | Improved |
| increase number of calibration folds E from 2 to 5 (ablation) | MMRL+NeRP · 10-dataset average, fine grid | Accuracy: 76.88 -> 77.32 | Improved |
| increase number of confusable pairs K (ablation) | NeRP · 10 datasets | Accuracy | Mixed |
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.
- MMRL — MMRL: Multi-Modal Representation Learning for Vision-Language Models (CVPR 2025) · better on 5 of 5 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
- Language Models as Black-Box Optimizers for Vision-Language Models (CVPR 2024 · similarity 0.89)
- A Closer Look at the Few-Shot Adaptation of Large Vision-Language Models (CVPR 2024 · similarity 0.89)
- ArGue: Attribute-Guided Prompt Tuning for Vision-Language Models (CVPR 2024 · similarity 0.89)
- PromptKD: Unsupervised Prompt Distillation for Vision-Language Models (CVPR 2024 · similarity 0.89)
- LaViP: Language-Grounded Visual Prompting (AAAI 2024 · similarity 0.88)
- JoAPR: Cleaning the Lens of Prompt Learning for Vision-Language Models (CVPR 2024 · similarity 0.88)
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.
