AI Paper Atlas

Research map / Video / Video-language understanding

Motion Question Answering via Modular Motion Programs

Mark Endo, Joy Hsu, Jiaman Li, Jiajun Wu

ICML 2023 · Video › Video-language understanding · approach: Video question answering · 4 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
Use modular neuro-symbolic motion programs instead of end-to-end or text-only predictionNS-Pose on BABEL-QA · BABEL-QA testaccuracy: 0.578; improvement over CLIP by 0.161, over MotionCLIP-MLP by 0.148, and over MotionCLIP-RNN by 0.158Improved
Add overlap between motion segments (ablation)NS-Pose · BABEL-QA testaccuracy: 0.540 -> 0.578; 0.038Improved
Use weakly supervised segment grounding instead of ground-truth action boundaries (ablation)NS-Pose · BABEL-QA testaccuracy: 0.553 -> 0.578; 0.025Improved
Replace dilated Conv1D temporal relation operators with a linear layer (ablation)NS-Pose · BABEL-QA testaccuracy: 0.578 vs 0.540No clear effect

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