Research map / Video / Video-language understanding
Expectation-Maximization Contrastive Learning for Compact Video-and-Language Representations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 EMCL to CLIP4Clip | CLIP4Clip · MSR-VTT Text-to-Video | R@1: 43.4 -> 46.9 | Helped |
| Add EMCL as a no-training inference module to CLIP4Clip | CLIP4Clip · MSR-VTT Video-to-Text | R@1: 42.4 -> 45.0 | Helped |
| Add EMCL with inverted softmax | EMCL-Net · MSR-VTT Text-to-Video | R@1: 46.8 -> 51.6 | Helped |
| Add EMCL with inverted softmax | EMCL-Net · ActivityNet Text-to-Video | R@1: 41.2 -> 50.6 | Helped |
| Add EMCL with inverted softmax | EMCL-Net · LSMDC Video-to-Text | R@1: 22.2 -> 26.7 | Helped |
| Use maintained parameter initialization instead of random initialization (ablation) | EMCL · MSR-VTT Text-to-Video | R@1: 43.3 -> 46.8 | Helped |
| Use maintained parameter initialization instead of random initialization (ablation) | EMCL · MSR-VTT Video-to-Text | R@1: 45.1 -> 46.5 | Helped |
| Vary the number of subspaces K (ablation) | EMCL · MSR-VTT Text-to-Video | R@1 | Mixed |
| Increase the EM iteration count until convergence (ablation) | EMCL · MSR-VTT Text-to-Video | R@1 | Helped |
| Tune sigma to 0.7 rather than the default sigma of 1 (ablation) | EMCL · MSR-VTT Text-to-Video | R@1: 43.8% to 51.6% | Helped |
| Add EMCL to the video captioning model | Base model_cap · MSR-VTT Video Captioning | METEOR: 29.8 -> 30.2 | Helped |
| Add EMCL to the video question answering model | Base model_qa · MSRVTT-QA | Accuracy: 45.0 -> 45.8 | Helped |
Closest papers by meaning
- Towards Efficient and Effective Text-to-Video Retrieval with Coarse-to-Fine Visual Representation Learning (AAAI 2024 · similarity 0.86)
- Ambiguity-Restrained Text-Video Representation Learning for Partially Relevant Video Retrieval (AAAI 2025 · similarity 0.86)
- Diffusion-Inspired Truncated Sampler for Text-Video Retrieval (NeurIPS 2024 · similarity 0.86)
- A Statistical Theory of Contrastive Learning via Approximate Sufficient Statistics (NeurIPS 2025 · similarity 0.86)
- Retrieval-Enhanced Contrastive Vision-Text Models (ICLR 2024 · similarity 0.86)
- Stitching Segments and Sentences towards Generalization in Video-Text Pre-training (AAAI 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.