Research map / Video / Video-language understanding
Knowing Where to Focus: Event-aware Transformer for Video Grounding
Jinhyun Jang, Jungin Park, Jin Kim, Hyeongjun Kwon, Kwanghoon Sohn
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 event reasoning with slot attention (ablation) | EaTR · QVHighlights val | R1@0.5: 55.10 -> 57.71 | Improved |
| add event localization loss (ablation) | EaTR · QVHighlights val | R1@0.5: 57.71 -> 59.10 | Improved |
| add gated fusion transformer layer (ablation) | EaTR · QVHighlights val | R1@0.5: 59.10 -> 61.36 | Improved |
| use gated fusion instead of addition or concatenation (ablation) | GF layer · QVHighlights val | R1@0.5: 58.68 -> 61.36 | Improved |
| add gate on top of MHCA fusion (ablation) | GF layer · QVHighlights val | R1@0.5: 59.74 -> 61.36 | Improved |
| use slot attention instead of cross-attention for event reasoning (ablation) | EaTR · QVHighlights val | R1@0.5: 57.35 -> 61.36 | Improved |
| increase slot attention iterations beyond K=3 (ablation) | EaTR · QVHighlights val | R1@0.5: 61.36 -> 59.16 (K=5) | Hurt |
| increase transformer layers from 3 to 5 (ablation) | EaTR · QVHighlights val | R1@0.5: 61.36 -> 61.35 | No clear effect |
| increase number of moment queries beyond 10 | EaTR · QVHighlights val | mAP | Hurt |
| add all EaTR components (ablation) | DETR baseline · ActivityNet Captions val_2 | R1@0.7: 31.74 -> 37.64 | Improved |
| add event reasoning only (ablation) | DETR baseline · Charades-STA test | R1@0.5: 66.75 -> 66.91 | No clear effect |
| event-aware dynamic moment queries instead of input-agnostic queries | Moment-DETR · QVHighlights val | highlight detection mAP: 37.15 vs QD-DETR 38.94 | Mixed |
Closest papers by meaning
- Decoupling Moment from Event for Video Temporal Grounding (ECCV 2026 · similarity 0.93)
- Training-free Video Temporal Grounding using Large-scale Pre-trained Models (ECCV 2024 · similarity 0.90)
- DART: Difficulty-Adaptive Routing for Zero-Shot Video Temporal Grounding (ECCV 2026 · similarity 0.90)
- Knowing Your Target: Target-Aware Transformer Makes Better Spatio-Temporal Video Grounding (ICLR 2025 · similarity 0.88)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · similarity 0.88)
- Talk2Event: Grounded Understanding of Dynamic Scenes from Event Cameras (NeurIPS 2025 · 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.
