Research map / Vision-language and multimodal / CLIP and zero-shot
DALIP: Distribution Alignment-based Language-Image Pre-Training for Domain-Specific Data
Junjie Wu, Jiangtao Xie, Zhaolin Zhang, Qilong Wang, Qinghua Hu, Peihua Li, Sen Xu
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 |
|---|---|---|---|
| replace [cls] matching with feature distribution alignment (DALIP) on TOL-1M | BioCLIP · 10 bio zero-shot tasks | mean accuracy: 33.0 -> 36.8 | Improved |
| DALIP tuned on TOL-10M | BioCLIP · 10 bio zero-shot tasks | mean accuracy: 38.7 -> 46.3 | Improved |
| DALIP on TOL-10M vs ArborCLIP-O on 40M data | ArborCLIP-O · 10 bio zero-shot tasks | mean accuracy: 34.4 -> 46.3 | Improved |
| DALIP vs OpenCLIP tuned on PlantMix-1.3M | OpenCLIP · General+plant zero-shot | mean accuracy: 40.1 -> 50.0 | Improved |
| DALIP-SigLIP vs SigLIP/ProLIP on PlantMix-1.3M | SigLIP · General+plant zero-shot | mean accuracy: 57.7 -> 62.4 | Improved |
| DALIP vs OpenCLIP ViT-L on PlantMix-13M | OpenCLIP ViT-L/14 · General+plant zero-shot | mean accuracy: 57.1 -> 60.5 | Improved |
| DALIP vs FLAIR on PlantMix-13M | FLAIR · General+plant zero-shot | mean accuracy: 45.1 -> 56.4 | Improved |
| DALIP on PlantMix-13M vs original OpenCLIP in general domain | OpenCLIP · General domain zero-shot | accuracy: 48.3 mean for original OpenCLIP vs 56.4 overall | Hurt |
| add 1st-order statistics to 2nd-order only (ablation) | DALIP · PlantMix-1.3M plant | plant mean accuracy: 44.8 -> 49.3 | Improved |
| add 2nd-order statistics to 1st-order only (ablation) | DALIP · PlantMix-1.3M IN-1K | accuracy: 30.7 -> 36.8 | Improved |
| use MBDC instead of BDC (ablation) | DALIP · PlantMix-1.3M plant | plant mean accuracy: 46.9 -> 49.3 | Improved |
| use MBDC instead of Moment Probing (ablation) | DALIP · PlantMix-1.3M plant | plant mean accuracy: 43.6 -> 49.3 | 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.
- BioCLIP — BioCLIP: A Vision Foundation Model for the Tree of Life (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
- BioCLIP 2: Emergent Properties from Scaling Hierarchical Contrastive Learning (NeurIPS 2025 · similarity 0.85)
- BiomedCoOp: Learning to Prompt for Biomedical Vision-Language Models (CVPR 2025 · similarity 0.84)
- Demystifying CLIP Data (ICLR 2024 · similarity 0.84)
- BIOMEDICA: An Open Biomedical Image-Caption Archive, Dataset, and Vision-Language Models Derived from Scientific Literature (CVPR 2025 · similarity 0.84)
- VLM4Bio: A Benchmark Dataset to Evaluate Pretrained Vision-Language Models for Trait Discovery from Biological Images (NeurIPS 2024 · similarity 0.84)
- CLIPCEIL: Domain Generalization through CLIP via Channel rEfinement and Image-text aLignment (NeurIPS 2024 · similarity 0.84)
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.
