Research map / Video / Video-language understanding
"GEB+: A Benchmark for Generic Event Boundary Captioning, Grounding and Retrieval"
Yuxuan Wang, Difei Gao, Licheng Yu, Weixian Lei, Matt Feiszli, Mike Zheng Shou
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 TPD modeling instead of no difference modeling (ablation) | ActBERT-revised (ResNet-roi+TSN) · Kinetic-GEB+ Boundary Captioning | CIDEr avg: 67.38 -> 74.71 | Improved |
| use TPD modeling instead of simple subtraction (ablation) | ActBERT-revised · Kinetic-GEB+ Boundary Captioning | CIDEr avg: 67.75 -> 74.71 | Improved |
| add simple subtraction difference modeling (ablation) | ActBERT-revised · Kinetic-GEB+ Boundary Captioning | CIDEr avg: 67.38 -> 67.75 | No clear effect |
| add TSN event-granularity feature to ResNet-roi (ablation) | ActBERT-revised · Kinetic-GEB+ Boundary Captioning | CIDEr avg: 51.93 -> 74.71 | Improved |
| add ResNet-roi instant feature to TSN (ablation) | ActBERT-revised · Kinetic-GEB+ Boundary Captioning | CIDEr avg: 70.80 -> 74.71 | Improved |
| add ResNet-conv to ResNet-roi + TSN (ablation) | ActBERT-revised · Kinetic-GEB+ Boundary Captioning | CIDEr avg: 74.71 -> 65.83 | Hurt |
| add ResNet-roi and ResNet-conv to TSN (ablation) | ActBERT-revised · Kinetic-GEB+ Boundary Captioning | CIDEr avg: 70.80 -> 65.83 | Hurt |
| add ResNet-conv to ResNet-roi (ablation) | ActBERT-revised · Kinetic-GEB+ Boundary Captioning | CIDEr avg: 51.93 -> 56.64 | Improved |
| use one-stream ActBERT-revised instead of two-stream UniVL-revised | TPD, ResNet-roi+TSN · Kinetic-GEB+ Boundary Captioning | CIDEr avg: 65.74 -> 74.71 | Improved |
| use FROZEN-revised instead of ActBERT-revised | Boundary Grounding · Kinetic-GEB+ Boundary Grounding | F1 avg: 30.66 -> 33.35 | Improved |
| use GEBD-based candidate sampling instead of default sampling | FROZEN-revised · Kinetic-GEB+ Boundary Grounding | F1 avg: 33.35 -> 33.23 | No clear effect |
| add difference modeling (ablation) | FROZEN-revised · Kinetic-GEB+ Boundary Caption-Video Retrieval | mAP: 22.44 -> 23.39 | Improved |
Closest papers by meaning
- Event-Equalized Dense Video Captioning (CVPR 2025 · similarity 0.88)
- Progress-Aware Video Frame Captioning (CVPR 2025 · similarity 0.87)
- E.T. Bench: Towards Open-Ended Event-Level Video-Language Understanding (NeurIPS 2024 · similarity 0.87)
- DELTAVID: Enhancing Fine-Grained Spatiotemporal Perception with Cross-Video Differences (NeurIPS 2026 · similarity 0.86)
- HourVideo: 1-Hour Video-Language Understanding (NeurIPS 2024 · similarity 0.86)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 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.
