Research map / Video / Video-language understanding
Learning from Inside: Self-driven Siamese Sampling and Reasoning for 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 siamese knowledge reasoning (soft-label self-distillation over clips) (ablation) | ClipBERT baseline · MSVD-QA | top-1 accuracy: 39.4 -> 44.7 | Helped |
| add siamese knowledge reasoning (soft-label self-distillation over clips) (ablation) | ClipBERT baseline · How2QA | top-1 accuracy: 79.1 -> 83.0 | Helped |
| add siamese knowledge generation on top of SKR (replace averaging) (ablation) | SiaSamRea · MSVD-QA | top-1 accuracy: 44.7 -> 45.5 | Helped |
| add siamese knowledge generation on top of SKR (replace averaging) (ablation) | SiaSamRea · How2QA | top-1 accuracy: 83.0 -> 84.1 | Helped |
| increase siamese clips from 1 to 12 (ablation) | SiaSamRea · MSVD-QA | top-1 accuracy: 39.9 -> 45.7 | Helped |
| increase siamese clips from 1 to 12 (ablation) | SiaSamRea · How2QA | top-1 accuracy: 79.6 -> 84.4 | Helped |
| increase siamese clips from 8 to 12 (ablation) | SiaSamRea · MSVD-QA | top-1 accuracy: 45.5 -> 45.7 | No clear effect |
| add learnable W1 in soft-label fusion (ablation) | SiaSamRea · MSVD-QA | top-1 accuracy: 44.9 -> 45.5 | Helped |
| add learnable W2 in soft-label fusion (ablation) | SiaSamRea · MSVD-QA | top-1 accuracy: 45.2 -> 45.5 | Helped |
| use real siamese clips instead of 8 copies of the anchor clip (ablation) | SiaSamRea · MSVD-QA | top-1 accuracy: 41.3 -> 45.5 | Helped |
| siamese sampling and reasoning vs ClipBERT* fine-tuned on filtered data | ClipBERT · TGIF-QA Action | accuracy: 75.1 -> 79.7 | Helped |
Closest papers by meaning
- VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection (CVPR 2025 · similarity 0.89)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.89)
- Modularized Self-Reflected Video Reasoner for Multimodal LLM with Application to Video Question Answering (ICML 2025 · similarity 0.88)
- Commonsense Video Question Answering through Video-Grounded Entailment Tree Reasoning (CVPR 2025 · similarity 0.88)
- Video-Context Aligned Transformer for Video Question Answering (AAAI 2024 · similarity 0.88)
- 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.