Research map / Video / Action recognition
What Changed and What Could Have Changed? State-Change Counterfactuals for Procedure-Aware Video Representation Learning
Chi-Hsi Kung, Frangil Ramirez, Juhyung Ha, Yi-Ting Chen, David Crandall, Yi-Hsuan Tsai
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 Before & After state-change text supervision (ablation) | Weak HierVL · GTEA temporal action segmentation | Acc: 77.3 -> 78.4 | Improved |
| add Before & After state-change text supervision (ablation) | Weak HierVL · EgoPER error detection | EDA: 71.2 -> 72.1 | Improved |
| add clip-level state-change counterfactuals (SC-CF) (ablation) | Before & After model · GTEA temporal action segmentation | Acc: 78.4 -> 79.1 | Improved |
| add clip-level state-change counterfactuals (SC-CF) (ablation) | Before & After model · EgoPER error detection | EDA: 72.1 -> 71.5 | Hurt |
| add only one type of video-level counterfactual (ID 4) (ablation) | Before & After + SC-CF · GTEA temporal action segmentation | Acc: 79.1 -> 78.7 | Hurt |
| add only one type of video-level counterfactual (ID 5) (ablation) | Before & After + SC-CF · GTEA temporal action segmentation | Acc: 79.1 -> 78.8 | Hurt |
| add one video-level counterfactual type (ID 4) (ablation) | Before & After + SC-CF · EgoPER error detection | EDA: 71.5 -> 72.2 | Improved |
| add the second video-level counterfactual type so both K-CF and M-CF are used (ablation) | Before & After + SC-CF + one video CF · GTEA temporal action segmentation | Acc: 78.7 -> 79.6 | Improved |
| add the second video-level counterfactual type so both K-CF and M-CF are used (ablation) | Before & After + SC-CF + one video CF · EgoPER error detection | EDA: 72.2 -> 72.3 | Improved |
| add all state changes and counterfactuals to HierVL video encoder (ablation) | Weak HierVL · GTEA temporal action segmentation | Acc: 77.3 -> 79.6 | Improved |
| add all state changes and counterfactuals to HierVL video encoder (ablation) | Weak HierVL · EgoPER error detection | EDA: 71.2 -> 72.3 | Improved |
Closest papers by meaning
- SCHEMA: State CHangEs MAtter for Procedure Planning in Instructional Videos (ICLR 2024 · similarity 0.88)
- Learning Procedural-Aware Video Representations Through State-Grounded Hierarchy Unfolding (AAAI 2026 · similarity 0.88)
- Change3D: Revisiting Change Detection and Captioning from A Video Modeling Perspective (CVPR 2025 · similarity 0.87)
- Learning Object State Changes in Videos: An Open-World Perspective (CVPR 2024 · similarity 0.87)
- Masked Temporal Interpolation Diffusion for Procedure Planning in Instructional Videos (ICLR 2025 · similarity 0.85)
- CofactVLA: Deconfounding Vision-Language-Action Models via Counterfactual Intervention (NeurIPS 2026 · 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.
