Research map / Video / Video-language understanding
Looking Beyond Visible Cues: Implicit Video Question Answering via Dual-Clue Reasoning
Tieyuan Chen, Huabin Liu, Yi Long Wang, Chaofan Gan, Mingxi Lv, Ziran Qin, Shijie Li, Liquan Shen, Junhui Hou, Zheng Wang, Weiyao Lin
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 dual context clues (action and intent) versus no guidance (ablation) | VideoChat2 · I-VQA | accuracy: improvements of 3.20% | Improved |
| add dual context clues versus Grounding-Dino object tracking (ablation) | VideoChat2 · I-VQA | accuracy: improvements of 2.53% | Improved |
| add dual context clues versus scene-graph triplet context (ablation) | VideoChat2 · I-VQA | accuracy: improvements of 2.36% | Improved |
| combine AIM and VEM modules instead of either alone (ablation) | IRM · I-VQA | accuracy | Improved |
| use IRM instead of fine-tuned VideoChat2 baseline | VideoChat2 · I-VQA | accuracy: improves accuracy by 4.87% | Improved |
| fine-tune open-source MLLMs on implicit questions without clue guidance | PLLaVA, Qwen2.5-VL, VideoChat2 · I-VQA | accuracy | No clear effect |
| iterate clue refinement 1-3 extra times (ablation) | IRM · I-VQA | accuracy: 55.14 -> 56.04 | Improved |
| use iterative clue enhancement instead of majority voting at same FLOPs | IRM · I-VQA | accuracy: 55.87 -> 56.04 | Improved |
| add noisy clues (50% extra) at inference (ablation) | VideoChat2 vs IRM · I-VQA | accuracy: drops by 5.36% (baseline) vs 1.31% (IRM) | Hurt |
| use IRM instead of GPT-4o on How questions | GPT-4o · I-VQA | accuracy: outperforms GPT-4o by 1.68% | Improved |
| use IRM instead of VideoChat2 zero-shot on PSAV | VideoChat2 · PSAV | recall: outperforming by 24.49% in recall and 6.86% in accuracy | Improved |
| use IRM instead of VideoChat2 on SUTD-Traffic event forecasting | VideoChat2 · SUTD-Traffic | accuracy: surpassing VideoChat2 and GPT-4o by 4.18% and 2.11% | Improved |
Closest papers by meaning
- Language-aware Visual Semantic Distillation for Video Question Answering (CVPR 2024 · similarity 0.90)
- Commonsense Video Question Answering through Video-Grounded Entailment Tree Reasoning (CVPR 2025 · similarity 0.90)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.90)
- Track the Answer: Extending TextVQA from Image to Video with Spatio-Temporal Clues (AAAI 2025 · similarity 0.89)
- VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection (CVPR 2025 · similarity 0.89)
- Not Another Text Benchmark: Putting the “Visual" Back in Visual Question Answering for Large Video Models (NeurIPS 2026 · similarity 0.89)
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.
