Research map / Vision-language and multimodal / CLIP and zero-shot
Cluster-Aware Neural Collapse Prompt Tuning for Long-Tailed Generalization of Vision-Language Models
Boyang Guo, Liang Li, Lin Peng, Yuhan Gao, Xichun Sheng, Chenggang Yan
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 L_TETF (ablation) | IVLP baseline · Base-to-New (tau=0.25) | harmonic mean: 70.35 -> 72.03 | Improved |
| add L_CC on top of L_TETF (ablation) | CPT · Base-to-New (tau=0.25) | harmonic mean: 72.03 -> 73.22 | Improved |
| add L_RS on top of L_TETF and L_CC (ablation) | CPT · Base-to-New (tau=0.25) | harmonic mean: 73.22 -> 73.58 | Improved |
| add L_RS on top of L_TETF (ablation) | CPT · Base-to-New (tau=0.25) | harmonic mean: 72.03 -> 72.42 | Improved |
| add all three losses (ablation) | IVLP baseline · Domain generalization (tau=0.25) | accuracy: 57.99 -> 60.16 | Improved |
| add all three losses (ablation) | IVLP baseline · Base-to-New (tau=0.06) | harmonic mean: 69.99 -> 72.47 | Improved |
| add L_CC on top of L_TETF (ablation) | CPT · Base-to-New (tau=0.06) | harmonic mean: 71.10 -> 71.85 | Improved |
| cosine K-means instead of Euclidean K-means (ablation) | CPT · ImageNet base-to-new (tau=0.25) | harmonic mean: 72.19 -> 72.62 | Improved |
| cosine K-means instead of spectral clustering (ablation) | CPT · ImageNet base-to-new (tau=0.25) | harmonic mean: 71.72 -> 72.62 | Improved |
| cosine K-means instead of K-medoids (ablation) | CPT · ImageNet base-to-new (tau=0.25) | harmonic mean: 71.86 -> 72.62 | Improved |
| cosine K-means instead of Euclidean K-means (ablation) | CPT · ImageNet base-to-new (tau=0.06) | seed std: 0.66 -> 0.52 | Improved |
| clusters of 16 prototypes (ablation) | CPT · ImageNet base-to-new | harmonic mean: 73.92 | Improved |
Closest papers by meaning
- Learning to Learn Better Visual Prompts (AAAI 2024 · similarity 0.89)
- Amend to Alignment: Decoupled Prompt Tuning for Mitigating Spurious Correlation in Vision-Language Models (ICML 2024 · similarity 0.88)
- ProAPO: Progressively Automatic Prompt Optimization for Visual Classification (CVPR 2025 · similarity 0.88)
- DPC: Dual-Prompt Collaboration for Tuning Vision-Language Models (CVPR 2025 · similarity 0.88)
- Frustratingly Easy Test-Time Adaptation of Vision-Language Models (NeurIPS 2024 · similarity 0.87)
- ZIP: An Efficient Zeroth-order Prompt Tuning for Black-box Vision-Language Models (ICLR 2025 · 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.
