Research map / Vision-language and multimodal / CLIP and zero-shot
AlignedNorm: Prompting Vision–Language Models via Coupled Prompt Field
Qi Ma, Chen-Yang Wang, Dehong Gao, Deng-Ping Fan
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 post-projection norm alignment L_proj (ablation) | MMRL++ · Base-to-new (11 datasets avg) | HM: 80.71 -> 80.97 | Improved |
| add layer-wise token norm alignment L_token alone (ablation) | MMRL++ · Base-to-new (11 datasets avg) | HM: 80.71 -> 80.52 | Hurt |
| add both L_proj and L_token (ablation) | MMRL++ · Base-to-new (11 datasets avg) | HM: 80.71 -> 81.45 | Improved |
| use L1 loss for norm alignment instead of hinge loss (ablation) | norm alignment · Base-to-new (11 datasets avg) | HM: 80.55 -> 80.97 | Improved |
| use L1 loss instead of smooth L1 or Ring loss (ablation) | norm alignment · Base-to-new (11 datasets avg) | HM: 80.85/80.90 -> 80.97 | Improved |
| add AlignedNorm to MMRL++ for new-class accuracy | MMRL++ · Base-to-new (11 datasets avg), new classes | accuracy (%): 76.48 -> 77.79 | Improved |
| add AlignedNorm (task-agnostic inference) vs MMRL++ with decoupled inference | MMRL++ · Base-to-new (11 datasets avg) | HM: 81.43 -> 81.45 | No clear effect |
| add AlignedNorm to MMRL++ w/o decoupling in cross-dataset transfer | MMRL++ · Cross-dataset (avg) | accuracy (%): 67.47 -> 67.61 | Improved |
| add AlignedNorm to MMRL++ in low-shot regime | MMRL++ · Few-shot, 1 shot | accuracy (%): 72.57 -> 72.93 | Improved |
| add AlignedNorm to MMRL++ at 16 shots | MMRL++ · Few-shot, 16 shots | accuracy (%): 84.07 -> 84.08 | No clear effect |
| add AlignedNorm to MMRL++ w/o decoupling in domain generalization | MMRL++ · Domain generalization (avg) | accuracy (%): 60.93 -> 60.96 | No clear effect |
| add AlignedNorm training loss | MMRL++ · training cost (ImageNet) | training time (s): 0.357 -> 0.357 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MMRL++ — MMRL: Multi-Modal Representation Learning for Vision-Language Models (CVPR 2025) · better on 7 of 8 shared results
- MaPLe — MaPLe: Multi-Modal Prompt Learning (CVPR 2023) · 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
- Language Models as Black-Box Optimizers for Vision-Language Models (CVPR 2024 · similarity 0.92)
- IPO: Interpretable Prompt Optimization for Vision-Language Models (NeurIPS 2024 · similarity 0.91)
- DPC: Dual-Prompt Collaboration for Tuning Vision-Language Models (CVPR 2025 · similarity 0.91)
- CoPL: Contextual Prompt Learning for Vision-Language Understanding (AAAI 2024 · similarity 0.91)
- Learning to Learn Better Visual Prompts (AAAI 2024 · similarity 0.90)
- Amend to Alignment: Decoupled Prompt Tuning for Mitigating Spurious Correlation in Vision-Language Models (ICML 2024 · 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.
