Research map / Video / Video-language understanding
MultiVENT: Multilingual Videos of Events and Aligned Natural Text
Kate Sanders, David Etter, Reno Kriz, Benjamin Van Durme
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 multilingual pretraining (XLM-R text encoder, LAION-5B) to pooled CLIP | pooled CLIP · MultiVENT non-English | Recall@1: 6.2 -> 32.6 | Improved |
| add multilingual pretraining to pooled CLIP | pooled CLIP · MultiVENT non-English | Recall@10: 22.4 -> 79.5 | Improved |
| add multilingual pretraining to pooled CLIP | pooled CLIP · MultiVENT non-English | Precision@10 (event description): 58.2 -> 61.5 | Improved |
| add multilingual pretraining to pooled CLIP | pooled CLIP · MultiVENT English | Recall@1: 55.9 -> 55.9 | No clear effect |
| add multilingual pretraining to pooled CLIP | pooled CLIP · MultiVENT English | Recall@10: 91.3 -> 92.3 | No clear effect |
| evaluate on non-English instead of English videos | InternVideo · MultiVENT | Recall@1: 53.8 -> 5.7 | Hurt |
| evaluate on non-English instead of English videos | MultiCLIP · MultiVENT | Recall@1: 55.9 -> 32.6 | Hurt |
| evaluate on non-English instead of English videos | CLIP2Video · MultiVENT | Recall@1: 41.3 -> 2.4 | Hurt |
| evaluate on non-English videos with English event queries | MultiCLIP · MultiVENT | Precision@10: 80.6 -> 61.5 | Hurt |
| use news broadcast instead of raw footage videos | annotation · MultiVENT disaster videos | entities per video: 3.63 -> 13.78 | Improved |
| use news broadcast instead of edited footage | annotation · MultiVENT disaster videos | on-screen text entities per video: 2.36 -> 7.87 | Improved |
| use MultiCLIP instead of Clip2Video on MSR-VTT | Clip2Video · MSR-VTT | Recall@1: 29.8 -> 38.4 | Improved |
Closest papers by meaning
- MultiVENT 2.0: A Massive Multilingual Benchmark for Event-Centric Video Retrieval (CVPR 2025 · similarity 0.94)
- Beyond Grounding: Extracting Fine-Grained Event Hierarchies across Modalities (AAAI 2024 · similarity 0.87)
- Bridging Information Asymmetry in Text-video Retrieval: A Data-centric Approach (ICLR 2025 · similarity 0.85)
- EventSum: A Large-Scale Event-Centric Summarization Dataset for Chinese Multi-News Documents (AAAI 2025 · similarity 0.84)
- InternVid: A Large-scale Video-Text Dataset for Multimodal Understanding and Generation (ICLR 2024 · similarity 0.83)
- E.T. Bench: Towards Open-Ended Event-Level Video-Language Understanding (NeurIPS 2024 · similarity 0.83)
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.
