AI Paper Atlas

Research map / Video / Video-language understanding

VALUE: A Multi-Task Benchmark for Video-and-Language Understanding Evaluation

Linjie Li, Jie Lei, Zhe Gan, Licheng Yu, Yen-Chun Chen, Rohit Pillai, Yu Cheng, Luowei Zhou, Xin Eric Wang, William Yang Wang, Tamara Lee Berg, Mohit Bansal and 3 more

NeurIPS 2021 · Video › Video-language understanding · approach: Video-language pretraining · 41 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 subtitle channel to video-only input (ablation)HERO · VALUE valMeta-Ave: 39.67 -> 52.52Improved
add video channel to subtitle-only input (ablation)HERO · VALUE valMeta-Ave: 40.40 -> 52.52Improved
HERO cross-modal fusion instead of two-stream late fusion (ablation)HERO · VALUE valMeta-Ave: 48.66 -> 52.52Improved
HERO fusion instead of sequence concat (ablation)HERO · VALUE valMeta-Ave: 51.53 -> 52.52Mixed
CLIP-ViT 2D features instead of ResNet (ablation)HERO finetune-only · VALUE valMeta-Ave: 48.29 -> 53.40Improved
CLIP-ViT+SlowFast instead of ResNet+SlowFast (ablation)HERO finetune-only · VALUE valMeta-Ave: 52.52 -> 54.52Improved
add ResNet to S3D features (ablation)HERO finetune-only · VALUE valMeta-Ave: 52.78 -> 49.34Hurt
multi-task by task type vs single-taskHERO finetune-only · VALUE testMeta-Ave: 50.44 -> 52.00Improved
multi-task by domain vs single-taskHERO finetune-only · VALUE testMeta-Ave: 50.44 -> 51.97Improved
all-task training vs single-taskHERO finetune-only · VALUE testMeta-Ave: 50.44 -> 52.33Improved
finetune each task after all-task trainingHERO finetune-only · VALUE testMeta-Ave: 52.33 -> 52.89Improved
HERO pre-training before finetuning (single-task)HERO · VALUE testMeta-Ave: 50.44 -> 57.00Improved

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