Research map / Video / Video-language understanding
Hierarchical Conditional Relation Networks for Video Question Answering
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Model high-order temporal relations rather than restricting relation order to k_max=1 (ablation) | HCRN · TGIF-QA Action | accuracy: 65.2 -> 75.0 | Helped |
| Use full hierarchical CRN processing instead of a single video-level CRN (ablation) | HCRN · TGIF-QA Action | accuracy: 66.2 -> 75.1 | Helped |
| Use full hierarchical CRN processing instead of clip CRNs followed by pooling (ablation) | HCRN · TGIF-QA Transition | accuracy: 80.5 -> 81.2 | Helped |
| Include motion features and motion conditioning (ablation) | HCRN · TGIF-QA Count | MSE: 4.38 -> 3.88 | Helped |
| Include long-term motion conditioning (ablation) | HCRN · TGIF-QA Count | MSE: 3.92 -> 3.88 | Helped |
| Include question conditioning at both clip and video levels (ablation) | HCRN · TGIF-QA Action | accuracy: 66.5 -> 75.1 | Helped |
| Use self-gating for question-conditioned CRNs (ablation) | HCRN · TGIF-QA Action | accuracy: 74.1 -> 75.1 | Helped |
| Apply self-gating to motion-conditioned as well as question-conditioned CRNs (ablation) | HCRN · TGIF-QA Action | accuracy: 75.1 -> 73.3 | Hurt |
| Increase sampling resolution t while keeping k_max=n-1 (ablation) | HCRN · TGIF-QA Action | accuracy: 75.0; 75.1; 73.6; 75.4; 74.1 | No clear effect |
| Add a hierarchy level to group clips into sub-videos (ablation) | HCRN · MSRVTT-QA, 24 clips | accuracy: 35.6 -> 35.6 | No clear effect |
| Add a hierarchy level to group clips into sub-videos (ablation) | HCRN · MSRVTT-QA, 24 clips | run time: 4 times reduction | Helped |
Closest papers by meaning
- STAIR: Spatial-Temporal Reasoning with Auditable Intermediate Results for Video Question Answering (AAAI 2024 · similarity 0.89)
- MoReVQA: Exploring Modular Reasoning Models for Video Question Answering (CVPR 2024 · similarity 0.88)
- Commonsense Video Question Answering through Video-Grounded Entailment Tree Reasoning (CVPR 2025 · similarity 0.88)
- VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection (CVPR 2025 · similarity 0.88)
- Divide and Conquer: Exploring Language-centric Tree Reasoning for Video Question-Answering (ICML 2025 · similarity 0.88)
- Language-aware Visual Semantic Distillation for Video Question Answering (CVPR 2024 · similarity 0.87)
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.