Research map / Video / Video-language understanding
Towards Long-Form Video Understanding
Check 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 |
|---|---|---|---|
| Use object-centric long-range transformer modeling rather than short-term or frame-centric modeling | LVU models · LVU tasks (8 of 9) | task performance | Helped |
| Use object-centric modeling rather than frame-centric VideoBERT | LVU · LVU tasks | overall rank: 2.22 -> 1.33 | Helped |
| Add self-supervised pre-training (ablation) | Object Transformer · LVU relation | accuracy: 46.9 -> 54.7 | Helped |
| Add compatibility prediction to masked-instance pre-training (ablation) | Object Transformer · LVU relation | accuracy: 54.7 -> 50.0 | Hurt |
| Replace transformer with average pooling (ablation) | Object Transformer · LVU relation | accuracy: 54.7 -> 37.5 | Hurt |
| Replace transformer with max pooling (ablation) | Object Transformer · LVU relation | accuracy: 54.7 -> 50.0 | Hurt |
| Add non-person object features (ablation) | Person-feature Object Transformer · LVU relation | accuracy: 54.7 -> 54.7 | No clear effect |
| Increase pre-training videos from 10k to 30k (ablation) | Object Transformer · LVU relation | accuracy: 50.0 -> 54.7 | Helped |
| Late-fuse Object Transformer context with short-term predictions | AVA V2.2 action detection · AVA V2.2 | mAP: 29.4 -> 31.0 | Helped |
| Use masked context without late fusion (ablation) | AVA V2.2 Object Transformer · AVA V2.2 | mAP: 29.4 -> 29.3 | No clear effect |
| Add object-level modeling to writer prediction | Object Transformer · LVU writer | accuracy: 38.5 -> 34.5 | Hurt |
Closest papers by meaning
- Memory Consolidation Enables Long-Context Video Understanding (ICML 2024 · similarity 0.90)
- Re-thinking Temporal Search for Long-Form Video Understanding (CVPR 2025 · similarity 0.89)
- Compositional Video Understanding with Spatiotemporal Structure-based Transformers (CVPR 2024 · similarity 0.87)
- Understanding Long Videos with Multimodal Language Models (ICLR 2025 · similarity 0.87)
- Understanding Video Transformers via Universal Concept Discovery (CVPR 2024 · similarity 0.87)
- SEAL: Semantic Attention Learning for Long Video Representation (CVPR 2025 · 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.