Research map / Vision-language and multimodal / CLIP and zero-shot
Dual-Generalization-aware Minimization for Continual Fine-Tuning of Vision-Language Models
Yanan Chen, Tieliang Gong, Yanle Lyu, Yuanhong Zhang, Weizhan 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 DGM to Finetune | Finetune · CIFAR-100 B0_Inc10 | last accuracy: 79.30 -> 81.12 | Improved |
| add DGM to ZSCL | ZSCL · CIFAR-100 B0_Inc10 | last accuracy: 78.33 -> 80.45 | Improved |
| add DGM to MG-CLIP | MG-CLIP · CUB-200 B0_Inc20 | last accuracy: 65.72 -> 67.05 | Improved |
| add DGM to RAPF | RAPF · CUB-200 B0_Inc20 | last accuracy: 76.20 -> 76.34 | No clear effect |
| add DGM to RAPF | RAPF · CUB-200 B100_Inc20 | average accuracy: 79.47 -> 79.34 | Hurt |
| add DGM to Finetune | Finetune · ImageNet-1K B0_Inc100 | last accuracy: 71.83 -> 73.40 | Improved |
| add DGM to ZSCL | ZSCL · Zero-shot (Food101, ImageNet-100, ImageNet-1K avg) | average accuracy: 70.77 -> 75.90 | Improved |
| add DGM to RAPF | RAPF · Zero-shot (Food101, ImageNet-100, ImageNet-1K avg) | average accuracy: 65.79 -> 68.25 | Improved |
| add DGM to DMNSP | DMNSP · Zero-shot (Food101, ImageNet-100, ImageNet-1K avg) | average accuracy: 72.20 -> 73.46 | Improved |
| add zero-shot and noisy-visual perturbations to SAM (ablation) | Finetune · CIFAR-100 | last accuracy: 80.42 -> 81.48 | Improved |
| add zero-shot and noisy-visual perturbations to SAM (ablation) | Finetune · CUB-200 | last accuracy: 60.40 -> 61.80 | Improved |
| add zero-shot and noisy-visual perturbations to SAM (ablation) | Finetune · Tiny-ImageNet | last accuracy: 76.20 -> 77.05 | Improved |
Closest papers by meaning
- SharpZO: Hybrid Sharpness-Aware Vision Language Model Prompt Tuning via Forward-Only Passes (NeurIPS 2025 · similarity 0.90)
- Frustratingly Easy Test-Time Adaptation of Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- On the Test-Time Zero-Shot Generalization of Vision-Language Models: Do We Really Need Prompt Learning? (CVPR 2024 · similarity 0.90)
- A Closer Look at the Few-Shot Adaptation of Large Vision-Language Models (CVPR 2024 · similarity 0.89)
- Test-Time Adaptation with CLIP Reward for Zero-Shot Generalization in Vision-Language Models (ICLR 2024 · similarity 0.89)
- Synthetic Data is an Elegant GIFT for Continual Vision-Language Models (CVPR 2025 · similarity 0.89)
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.
