Research map / Video / Video-language understanding
Knowledge-Constrained Answer Generation for Open-Ended Video Question Answering
Read the paperCheck 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 |
|---|---|---|---|
| Add adaptive external knowledge and multi-stream information control to ClipBERT-based freeform generation | Open-QA · Open-QA | accuracy: 20.36 -> 27.04 | Helped |
| Add adaptive external knowledge and multi-stream information control | NExT-QA · NExT-QA | accuracy: 24.17 -> 28.21 | Helped |
| Use freeform GPT2 generation instead of a multi-label classifier (ablation) | KcGA on NExT-QA · NExT-QA | accuracy: 26.94 -> 28.21 | Helped |
| Use multi-stream information control instead of Cat1 concatenation (ablation) | KcGA · Open-QA | accuracy: 25.26 -> 27.04 | Helped |
| Use multi-stream information control instead of Cat2 concatenation (ablation) | KcGA · Open-QA | accuracy: 25.01 -> 27.04 | Helped |
| Use multi-stream information control instead of SRL fusion (ablation) | KcGA · Open-QA | accuracy: 26.11 -> 27.04 | Helped |
| Use multi-stream information control instead of SCA fusion (ablation) | KcGA · Open-QA | accuracy: 23.04 -> 27.04 | Helped |
| Use freeform generation instead of multi-label classification (ablation) | KcGA · Open-QA | accuracy: 22.97 -> 27.04 | Helped |
| Remove CLIP denoising of external knowledge (ablation) | KcGA · Open-QA | accuracy: 27.04 -> 24.91 | Hurt |
| Remove CLIP denoising of visual knowledge (ablation) | KcGA · Open-QA | accuracy: 27.04 -> 26.55 | Hurt |
| Remove memory computation over external knowledge (ablation) | KcGA · Open-QA | accuracy: 27.04 -> 26.46 | Hurt |
| Add memory computation over external knowledge (ablation) | KcGA · Open-QA out-of-domain answers | accuracy: 4.16 -> 4.55 | Helped |
Closest papers by meaning
- Streaming Video Question-Answering with In-context Video KV-Cache Retrieval (ICLR 2025 · similarity 0.88)
- Commonsense Video Question Answering through Video-Grounded Entailment Tree Reasoning (CVPR 2025 · similarity 0.88)
- Language-aware Visual Semantic Distillation for Video Question Answering (CVPR 2024 · similarity 0.88)
- Ranking Distillation for Open-Ended Video Question Answering with Insufficient Labels (CVPR 2024 · similarity 0.88)
- Separation of Powers: On Segregating Knowledge from Observation in LLM-enabled Knowledge-based Visual Question Answering (CVPR 2025 · similarity 0.87)
- MoReVQA: Exploring Modular Reasoning Models 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.