Research map / Video / Video-language understanding
Hierarchical Banzhaf Interaction for General Video-Language Representation Learning
Peng Jin, Hao Li, Li Yuan, Shuicheng Yan, Jie Chen
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 Banzhaf interaction loss (ablation) | baseline · MSRVTT text-to-video retrieval | R@1: 46.6 -> 47.4 | Improved |
| add Banzhaf interaction loss (ablation) | baseline · MSRVTT-QA | accuracy: 44.5 -> 45.8 | Improved |
| add deep supervision (ablation) | baseline · MSRVTT text-to-video retrieval | R@1: 46.6 -> 47.2 | Improved |
| add deep supervision (ablation) | baseline · MSRVTT-QA | accuracy: 44.5 -> 46.0 | Improved |
| add self-distillation on top of Banzhaf and deep supervision (ablation) | HBI · MSRVTT text-to-video retrieval | R@1: 48.2 -> 48.6 | Improved |
| add representation reconstruction (ablation) | HBI · MSRVTT text-to-video retrieval | R@1: 48.6 -> 49.4 | Improved |
| add representation reconstruction (ablation) | HBI · MSRVTT-QA | accuracy: 46.2 -> 46.4 | Improved |
| use CNN+SA prediction head instead of None (ablation) | HBI V2 · MSRVTT text-to-video retrieval | R@1: 47.7 -> 49.4 | Improved |
| use CNN+SA prediction head instead of MLP+SA (ablation) | HBI V2 · MSRVTT text-to-video retrieval | R@1: 48.4 -> 49.4 | Improved |
| distill entity level to both action and event levels instead of E->A only (ablation) | HBI V2 · MSRVTT text-to-video retrieval | R@1: 48.7 -> 49.4 | Improved |
| distill A->O in addition to E->A (ablation) | HBI V2 · MSRVTT text-to-video retrieval | R@1: 49.4 -> 48.8 | Hurt |
| use cluster numbers 6,2,16,4 instead of 9,3,16,4 (ablation) | HBI V2 · MSRVTT text-to-video retrieval | R@1: 48.6 -> 49.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- UATVR — UATVR: Uncertainty-Adaptive Text-Video Retrieval (ICCV 2023) · better on 2 of 2 shared results
- TS2-Net — TS2-Net: Token Shift and Selection Transformer for Text-Video Retrieval (ECCV 2022) · better on 1 of 1 shared result
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
- Video-Text As Game Players: Hierarchical Banzhaf Interaction for Cross-Modal Representation Learning (CVPR 2023 · similarity 0.95)
- InternVid: A Large-scale Video-Text Dataset for Multimodal Understanding and Generation (ICLR 2024 · similarity 0.87)
- Towards Efficient and Effective Text-to-Video Retrieval with Coarse-to-Fine Visual Representation Learning (AAAI 2024 · similarity 0.86)
- SF2T: Self-supervised Fragment Finetuning of Video-LLMs for Fine-Grained Understanding (CVPR 2025 · similarity 0.86)
- Modeling Multimodal Social Interactions: New Challenges and Baselines with Densely Aligned Representations (CVPR 2024 · similarity 0.85)
- Dissecting Multimodality in VideoQA Transformer Models by Impairing Modality Fusion (ICML 2024 · similarity 0.85)
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.
