Research map / Vision-language and multimodal / CLIP and zero-shot
ProLoG: Hybrid Prompt and LoRA Based Adaptation of Vision-Language Models for OOD Generalization
Jungwuk Park, Dong-Jun Han, Jaekyun Moon
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 the hybrid prompt+LoRA network (ablation) | CLIP · Base-to-new, 11 datasets (ViT-B/16) | harmonic mean: 71.70 -> 78.96 | Improved |
| add the hybrid network (ablation) | CLIP · Base-to-new, 11 datasets (ViT-B/16) | new-class accuracy: 74.22 -> 73.57 | No clear effect |
| add augmentation-based regularization loss (ablation) | hybrid network · Base-to-new, 11 datasets (ViT-B/16) | new-class accuracy: 73.57 -> 76.18 | Improved |
| add augmentation-based regularization loss (ablation) | hybrid network · Base-to-new, 11 datasets (ViT-B/16) | base-class accuracy: 85.21 -> 84.52 | Hurt |
| add augmentation-based regularization loss (ablation) | hybrid network · Base-to-new, 11 datasets (ViT-B/16) | harmonic mean: 78.96 -> 80.13 | Improved |
| add masking strategy for LoRA on class-specific tokens (ablation) | hybrid network + regularization · Base-to-new, 11 datasets (ViT-B/16) | harmonic mean: 80.13 -> 80.33 | Improved |
| add task-similarity adaptive inference (ablation) | ProLoG · Base-to-new, 11 datasets (ViT-B/16) | new-class accuracy: 76.32 -> 76.89 | Improved |
| add task-similarity adaptive inference (ablation) | ProLoG · Base-to-new, 11 datasets (ViT-B/16) | harmonic mean: 80.33 -> 80.64 | Improved |
| use ProLoG instead of PromptSRC | CLIP ViT-B/32 · Base-to-new, 11 datasets | harmonic mean: 75.95 -> 76.84 | Improved |
| use ProLoG instead of PromptSRC | CLIP ViT-B/32 · Base-to-new, 11 datasets | base-class accuracy: 81.28 -> 81.05 | No clear effect |
| use ProLoG instead of CoPrompt | CLIP ViT-B/16 · Cross-dataset (UCF101 base) | avg target accuracy: 60.26 -> 64.19 | Improved |
| use ProLoG instead of PromptSRC | CLIP ViT-B/16 · Cross-dataset (DTD base) | avg target accuracy: 65.69 -> 66.10 | Improved |
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.
- MaPLe — MaPLe: Multi-Modal Prompt Learning (CVPR 2023) · better on 3 of 3 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
- Overcoming the Pitfalls of Vision-Language Model Finetuning for OOD Generalization (ICLR 2024 · similarity 0.91)
- Frustratingly Easy Test-Time Adaptation of Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- Rethinking Few-Shot Adaptation of Vision-Language Models in Two Stages (CVPR 2025 · similarity 0.89)
- LLMind: Bio-inspired Training-free Adaptive Visual Representations for Vision-Language Models (CVPR 2026 · similarity 0.89)
- A Closer Look at the Few-Shot Adaptation of Large Vision-Language Models (CVPR 2024 · similarity 0.89)
- Language Models as Black-Box Optimizers for Vision-Language Models (CVPR 2024 · 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.
