Research map / Video / Video-language understanding
InstrAct: Towards Action-Centric Understanding in Instructional Videos
Zhuoyi Yang, Jiapeng Yu, Reuben Tan, Boyang Albert Li, Huijuan Xu
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 action-centric hard negatives (verb-altered/order-swapped) to training | CLIP-ViP · InstrAct-Semantic | R@1 (%): 18.4 -> 38.5 | Improved |
| add action-centric hard negatives | CLIP4Clip · InstrAct-Semantic | R@1 (%): 14.8 -> 44.7 | Improved |
| add Masked Action Modeling on top of Action Perceiver (ablation) | InstrAction (ViCLIP) · InstrAct-Semantic | R@1 (%): 40.7 -> 43.7 | Improved |
| add DTW-Align on top of Action Perceiver (ablation) | InstrAction (ViCLIP) · InstrAct-Logic | ACC (%): 37.2 -> 40.2 | Improved |
| add DTW-Align on top of Action Perceiver (ablation) | InstrAction (ViCLIP) · InstrAct-Semantic | R@1 (%): 40.7 -> 38.7 | Hurt |
| combine MAM and DTW-Align (ablation) | InstrAction (ViCLIP) · InstrAct-Logic | ACC (%): 37.2 -> 44.7 | Improved |
| full InstrAction framework over hard negatives alone | MIL-NCE · InstrAct-Logic | ACC (%): 6.1 -> 4.7 | No clear effect |
| full InstrAction over hard negatives only | CLIP-ViP · InstrAct-Logic | ACC (%): 30.2 -> 37.1 | Improved |
| full InstrAction over hard negatives only | CLIP4Clip · InstrAct-Semantic | R@1 (%): 44.7 -> 43.3 | Hurt |
| Action Perceiver latents instead of raw frame embeddings (ablation) | InstrAction · DTW alignment cost | normalized DTW cost | Improved |
| InstrAction vs strongest baseline | InstrAction · InstrAct-Dynamics | Mean R@1: 30.68 -> 34.53 | Improved |
Closest papers by meaning
- Masked Temporal Interpolation Diffusion for Procedure Planning in Instructional Videos (ICLR 2025 · similarity 0.89)
- Open-Vocabulary Video Relation Extraction (AAAI 2024 · similarity 0.87)
- Kronecker Mask and Interpretive Prompts are Language-Action Video Learners (ICLR 2025 · similarity 0.86)
- Storyboard-guided Alignment for Fine-grained Video Action Recognition (NeurIPS 2025 · similarity 0.86)
- Disentangled Concepts Speak Louder Than Words: Explainable Video Action Recognition (NeurIPS 2025 · similarity 0.86)
- FAVOR-Bench: A Comprehensive Benchmark for Fine-Grained Video Motion Understanding (NeurIPS 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.
