Research map / Vision-language and multimodal / CLIP and zero-shot
Preserve and Sculpt: Manifold-Aligned Fine-tuning of Vision-Language Models for Few-Shot Learning
Dexia Chen, Qianjie Zhu, Weibing Li, Yue Yu, Tong Zhang, Ruixuan 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 Manifold Alignment Regularization to cross-entropy (ablation) | MPS-Tuning · 11-dataset average (16-shot) | accuracy: 85.41 -> 86.44 | Improved |
| add Hierarchical Manifold Sculpting to cross-entropy (ablation) | MPS-Tuning · 11-dataset average (16-shot) | accuracy: 85.41 -> 86.20 | Improved |
| add HMS on top of MAR (ablation) | MPS-Tuning · 11-dataset average (16-shot) | accuracy: 86.44 -> 86.85 | Improved |
| add MAR on top of HMS (ablation) | MPS-Tuning · ImageNet (16-shot) | accuracy: 74.77 -> 75.60 | Improved |
| add global alignment alone to HMS baseline (ablation) | MAR · 11-dataset average (1-shot) | accuracy: 72.35 -> 73.42 | Improved |
| add local alignment alone to HMS baseline (ablation) | MAR · 11-dataset average (16-shot) | accuracy: 86.20 -> 86.67 | Improved |
| combine global and local alignment instead of local only (ablation) | MAR · 11-dataset average (1-shot) | accuracy: 72.82 -> 73.55 | Improved |
| use MAR instead of feature-level cosine consistency (ablation) | MPS-Tuning · 11-dataset average (16-shot) | accuracy: 86.07 -> 86.85 | Improved |
| use MAR instead of logit-level KL consistency (ablation) | MPS-Tuning · 11-dataset average (16-shot) | accuracy: 84.20 -> 86.85 | Improved |
| use feature-level L1 consistency instead of none (no regularizer) (ablation) | HMS baseline · 11-dataset average (16-shot) | accuracy: 86.20 -> 86.24 | No clear effect |
| apply HMS to earlier layers beyond the last two (ablation) | MPS-Tuning · 11-dataset average (16-shot) | accuracy | Hurt |
| raise HMS weight lambda2 from 0.1 to 2 (ablation) | MPS-Tuning · 11-dataset average (16-shot) | accuracy: 86.85 -> 73.82 | Hurt |
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 4 of 4 shared results
- AMU-Tuning — AMU-Tuning: Effective Logit Bias for CLIP-based Few-shot Learning (CVPR 2024) · 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
- Homology Consistency Constrained Efficient Tuning for Vision-Language Models (NeurIPS 2024 · similarity 0.89)
- Locality Alignment Improves Vision-Language Models (ICLR 2025 · similarity 0.88)
- Topology-Aware Representation Alignment for Semi-Supervised Vision-Language Learning (NeurIPS 2026 · similarity 0.87)
- Making Large Vision Language Models to Be Good Few-Shot Learners (AAAI 2025 · similarity 0.87)
- TPR: Topology-Preserving Reservoirs for Generalized Zero-Shot Learning (NeurIPS 2024 · similarity 0.87)
- Connecting the Dots: Collaborative Fine-tuning for Black-Box Vision-Language Models (ICML 2024 · similarity 0.87)
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.
