AI Paper Atlas

Research map / Video / Video-language understanding

Video-Text As Game Players: Hierarchical Banzhaf Interaction for Cross-Modal Representation Learning

Peng Jin, Jinfa Huang, Pengfei Xiong, Shangxuan Tian, Chang Liu, Xiangyang Ji, Li Yuan, Jie Chen

CVPR 2023 · Video › Video-language understanding · approach: Video-language pretraining · 71 citations (OpenAlex)

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 changedWhereMeasuredEffect
Add hierarchical Banzhaf interaction learningCLIP (ViT-B/32) · MSRVTT, text-to-videoR@1: 47.0 -> 48.6Improved
Add hierarchical Banzhaf interaction learningCLIP (ViT-B/32) · ActivityNet Captions, text-to-videoR@1: 41.2 -> 42.2Improved
Add hierarchical Banzhaf interaction learningCLIP (ViT-B/32) · DiDeMo, text-to-videoR@1: 42.8 -> 46.9Improved
Add hierarchical Banzhaf interaction learningCLIP (ViT-B/32) · MSRVTT, video-to-textR@1: 45.3 -> 46.8Improved
Add hierarchical Banzhaf interaction learningCLIP (ViT-B/32) · ActivityNet Captions, video-to-textR@1: 41.4 -> 42.4Improved
Add hierarchical Banzhaf interaction learningCLIP (ViT-B/32) · DiDeMo, video-to-textR@1: 41.4 -> 46.2Improved
Use three semantic levels instead of one (ablation)HBI · MSRVTT, text-to-videoR@1: 47.5 -> 48.6Improved
Use three semantic levels instead of two (ablation)HBI · MSRVTT, text-to-videoR@1: 48.1 -> 48.6Improved
Use a tiny learned estimator instead of sampling to estimate Banzhaf Interaction (ablation)HBI · MSRVTT subset, text-to-videoiteration time: 19.79 seconds per iteration to 3.14 seconds per iterationImproved
Use CNN+SA instead of MLP+SA in the prediction header (ablation)HBI · MSRVTT, text-to-videoR@1: 46.6 -> 48.6Improved
Use CNN+SA instead of MLP in the prediction header (ablation)HBI · MSRVTT, text-to-videoR@1: 47.2 -> 48.6Improved
Distill entity-level similarity to both action and event levels (ablation)HBI · MSRVTT, text-to-videoR@1: 48.2 -> 48.6Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy