Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
CWCL: Cross-Modal Transfer with Continuously Weighted Contrastive Loss
Rakshith Sharma Srinivasa, Jaejin Cho, Chouchang Yang, Yashas Malur Saidutta, Ching-Hua Lee, Yilin Shen, Hongxia Jin
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 |
|---|---|---|---|
| CWCL instead of LiT contrastive loss (12-layer transformer text) | ViT-L/16 image-text model · ImageNet zero-shot | accuracy: 66.84 -> 74.41 | Improved |
| CWCL instead of LiT (12-layer transformer text) | ViT-L/16 image-text model · ImageNet-A zero-shot | accuracy: 37.31 -> 45.58 | Improved |
| CWCL instead of LiT (12-layer transformer text) | ViT-L/16 image-text model · ObjectNet zero-shot | accuracy: 45.08 -> 50.5 | Improved |
| CWCL instead of LiT (BERT-large text) | ViT-L/16 + BERT-large · ImageNet zero-shot | accuracy: 71.2 -> 76.48 | Improved |
| CWCL instead of LiT (BERT-large text) | ViT-L/16 + BERT-large · ImageNet-A zero-shot | accuracy: 40.28 -> 47.27 | Improved |
| CWCL instead of LiT for retrieval | ViT-L/16 + BERT-large · MS-COCO I->T R@1 | recall: 34.58 -> 40.36 | Improved |
| CWCL instead of LiT for retrieval | ViT-L/16 + BERT-large · MS-COCO T->I R@10 | recall: 65.87 -> 66.06 | No clear effect |
| CWCL instead of CL for retrieval | image-text model · MS-COCO I->T R@1 | recall: 30.42 -> 35.10 | Improved |
| CWCL instead of CL for speech-to-intent | Whisper + RoBERTa+S · SLURP zero-shot | top-1 accuracy: 40.35 -> 63.80 | Improved |
| CWCL instead of CL for speech-to-intent | Whisper + BART+Y · STOP zero-shot | top-1 accuracy: 55.67 -> 80.99 | Improved |
| CWCL instead of CL for keyword spotting | Whisper + BART+Y · GSCV2 zero-shot | top-1 accuracy: 56.33 -> 88.81 | Improved |
| Pre-trained Whisper speech encoder instead of random initialization (ablation) | CWCL speech-text · SLURP | top-1 accuracy: 26.17 -> 53.12 | Improved |
Later papers that compared with this method
1 later accepted paper reports CWCL as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- How Molecules Impact Cells: Unlocking Contrastive PhenoMolecular Retrieval (NeurIPS 2024 · better on 2 of 2 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
- Connect, Collapse, Corrupt: Learning Cross-Modal Tasks with Uni-Modal Data (ICLR 2024 · similarity 0.89)
- Understanding Transferable Representation Learning and Zero-shot Transfer in CLIP (ICLR 2024 · similarity 0.88)
- Language-Driven Cross-Modal Classifier for Zero-Shot Multi-Label Image Recognition (ICML 2024 · similarity 0.87)
- Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion (ICLR 2025 · similarity 0.86)
- Mind the Gap Between Prototypes and Images in Cross-domain Finetuning (NeurIPS 2024 · similarity 0.86)
- On the Comparison between Multi-modal and Single-modal Contrastive Learning (NeurIPS 2024 · similarity 0.86)
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.
