Research map / Vision-language and multimodal / Cross-modal learning
Connecting Multi-modal Contrastive Representations
Zehan Wang, Yang Zhao, Xize Cheng, Haifeng Huang, Jiageng Liu, Aoxiong Yin, Li Tang, Linjun Li, Yongqi Wang, Ziang Zhang, Zhou Zhao
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 |
|---|---|---|---|
| connect CLIP and CLAP via overlapping text, without paired audio-visual training data | Audio-visual representations · AVE and Flickr-SoundNet retrieval | mAP: 4.11, 4.13, 4.57, 4.92 | Improved |
| connect CLIP and ULIP-2 through images | 3D-language representations · ModelNet40 | Top1: 74.0 -> 64.9 | Mixed |
| add text-text inter-MCR alignment (ablation) | C-MCR audio-visual retrieval · AVE A2I | mAP: 3.28 -> 4.11 | Improved |
| add pseudo audio-visual inter-MCR alignment (ablation) | C-MCR audio-visual retrieval · AVE A2I | mAP: 4.09 -> 4.11 | No clear effect |
| add both inter-MCR alignment losses (ablation) | C-MCR audio-visual retrieval · AVE A2I | mAP: 0.22 -> 4.11 | Improved |
| add intra-CLIP and intra-CLAP alignment (ablation) | C-MCR audio-visual retrieval · AVE A2I | mAP: 3.14 -> 4.11 | Improved |
| add Gaussian noise and re-normalization to embeddings (ablation) | C-MCR audio-visual retrieval · AVE A2I | mAP: 2.77 -> 4.11 | Improved |
| add re-normalization after noise (ablation) | C-MCR audio-visual retrieval · AVE A2I | mAP: 3.70 -> 4.11 | Improved |
| use softmax memory aggregation instead of argmax selection (ablation) | C-MCR audio-visual retrieval · AVE A2I | mAP: 4.01 -> 4.11 | Improved |
| use semantic memory selection rather than random memory selection (ablation) | C-MCR audio-visual retrieval · AVE A2I | mAP: 2.62 -> 4.11 | Improved |
| vary the Gaussian noise variance (ablation) | C-MCR audio-visual retrieval · AVE and Flickr-SoundNet | average mAP | No clear effect |
Closest papers by meaning
- Extending Multi-modal Contrastive Representations (NeurIPS 2024 · similarity 0.93)
- Connect, Collapse, Corrupt: Learning Cross-Modal Tasks with Uni-Modal Data (ICLR 2024 · similarity 0.90)
- MM-Mixing: Multi-Modal Mixing Alignment for 3D Understanding (AAAI 2025 · similarity 0.88)
- CLIP-Driven Open-Vocabulary 3D Scene Graph Generation via Cross-Modality Contrastive Learning (CVPR 2024 · similarity 0.88)
- Gramian Multimodal Representation Learning and Alignment (ICLR 2025 · similarity 0.87)
- Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion (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.
