Research map / Video / Video-language understanding
Glance and Focus: Memory Prompting for Multi-Event Video Question Answering
Ziyi Bai, Ruiping Wang, Xilin CHEN
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 |
|---|---|---|---|
| Use dynamic event memories as prompts and add multi-level attention (ablation) | Glance-Focus · STAR | Accuracy: 53.12 -> 53.94 | Improved |
| Use dynamic event memories as prompts and add multi-level attention (ablation) | Glance-Focus · EgoTaskQA | Accuracy: 42.70 -> 44.27 | Improved |
| Learn event memories instead of using a frozen action detector (ablation) | Glance-Focus · STAR | Accuracy: 49.85 -> 53.94 | Improved |
| Learn event memories instead of using a frozen action detector (ablation) | Glance-Focus · EgoTaskQA | Accuracy: 42.62 -> 44.27 | Improved |
| Add unsupervised semantic diversity loss L_cls (ablation) | GF(uns) · AGQA | Accuracy: 54.27 -> 54.59 | Improved |
| Add temporal overlap loss L_iou (ablation) | GF(uns) · AGQA | Accuracy: 52.89 -> 54.59 | Improved |
| Add individual certainty loss L_cert (ablation) | GF(uns) · AGQA | Accuracy: 54.12 -> 54.59 | Improved |
| Add supervised event classification loss L_cls (ablation) | GF(sup) · AGQA | Accuracy: 53.19 -> 55.08 | Improved |
| Add supervised temporal span L1 loss (ablation) | GF(sup) · AGQA | Accuracy: 54.26 -> 55.08 | Improved |
| Train event-memory generation without event labels rather than with supervision | Glance-Focus · STAR | Mean accuracy: 53.86 vs 53.94 | No clear effect |
| Train event-memory generation without event labels rather than with supervision | Glance-Focus · EgoTaskQA | All accuracy: 43.06 vs 44.27 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MIST — MIST: Multiple Instance Self-Training Framework for Video Anomaly Detection (CVPR 2021) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MoReVQA: Exploring Modular Reasoning Models for Video Question Answering (CVPR 2024 · similarity 0.89)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.89)
- Temporal Chain of Thought: Long-Video Understanding by Thinking in Frames (NeurIPS 2025 · similarity 0.88)
- BDIQA: A New Dataset for Video Question Answering to Explore Cognitive Reasoning through Theory of Mind (AAAI 2024 · similarity 0.88)
- ReXTime: A Benchmark Suite for Reasoning-Across-Time in Videos (NeurIPS 2024 · similarity 0.88)
- Enhancing Temporal Understanding in Video-LLMs through Stacked Temporal Attention in Vision Encoders (NeurIPS 2025 · similarity 0.88)
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.
