Research map / Video / Video-language understanding
NaQ: Leveraging Narrations As Queries To Supervise Episodic Memory
Santhosh Kumar Ramakrishnan, Ziad Al-Halah, Kristen Grauman
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 narration-derived queries to VSLNet training | VSLNet · Ego4D NLQ val | R@1, IoU=0.3: 5.45 -> 10.26 | Improved |
| Add narration-derived queries to EgoVLP training | EgoVLP · Ego4D NLQ val | R@1, IoU=0.3: 10.84 -> 15.90 | Improved |
| Add narration-derived queries to ReLER* training | ReLER* · Ego4D NLQ val | R@1, IoU=0.3: 14.66 -> 19.31 | Improved |
| Add Ego4D narration-derived query training | VSLNet · TACoS val | R@1, IoU=0.3: 20.10 -> 23.39 | Improved |
| Add Ego4D narration-derived query training | EgoVLP · TACoS val | R@1, IoU=0.3: 16.52 -> 18.24 | Improved |
| Add temporal response jittering to narration windows (ablation) | NaQ with VSLNet · Ego4D NLQ val | R@1, IoU=0.3: 9.80 -> 10.26 | Improved |
| Add temporal response jittering to narration windows (ablation) | NaQ with EgoVLP · Ego4D NLQ val | R@1, IoU=0.3: 15.25 -> 15.90 | Improved |
| Add temporal response jittering to narration windows (ablation) | NaQ with ReLER* · Ego4D NLQ val | R@1, IoU=0.3: 18.51 -> 19.31 | Improved |
| Add NLQ-only finetuning after joint NaQ and NLQ training (ablation) | VSLNet + NaQ · Ego4D NLQ val | R@1, IoU=0.3: 10.26 -> 9.34 | Hurt |
| Use temporal expansion scale S=5.0 rather than S=2.5 (ablation) | EgoVLP + NaQ · Ego4D NLQ val | R@1, IoU=0.5: 8.64 -> 9.46 | Improved |
| Train with NaQ queries for low-shot objects | VSLNet · Ego4D NLQ val, low-shot objects | R@1, IoU=0.3: 4.05 -> 10.30 | Improved |
| Train with NaQ queries for query template 'In what location did I see X?' | ReLER* · Ego4D NLQ val | R@1, IoU=0.5: 5.46 -> 4.78 | Hurt |
Closest papers by meaning
- Grounded Question-Answering in Long Egocentric Videos (CVPR 2024 · similarity 0.88)
- Omnia de EgoTempo: Benchmarking Temporal Understanding of Multi-Modal LLMs in Egocentric Videos (CVPR 2025 · similarity 0.87)
- Narrating the Video: Boosting Text-Video Retrieval via Comprehensive Utilization of Frame-Level Captions (CVPR 2025 · similarity 0.87)
- ReWind: Understanding Long Videos with Instructed Learnable Memory (CVPR 2025 · similarity 0.87)
- HourVideo: 1-Hour Video-Language Understanding (NeurIPS 2024 · similarity 0.86)
- EgoExoBench: A Benchmark for First- and Third-person View Video Understanding in MLLMs (NeurIPS 2025 · similarity 0.86)
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.
