Research map / Large language models / Agents, code and math
LLM-Powered Query Expansion for Enhancing Boundary Prediction in Language-Driven Action Localization
Zirui Shang, Xinxiao Wu, Shuo Yang
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 query-guided temporal modeling (QTM) (ablation) | QD-DETR · Charades-STA | R1@0.5: 58.17 -> 59.06 | Improved |
| add boundary probability modeling (BPM) (ablation) | QD-DETR · Charades-STA | R1@0.5: 58.17 -> 59.33 | Improved |
| add both QTM and BPM (ablation) | QD-DETR · TACoS | R1@0.7: 20.27 -> 24.04 | Improved |
| add QTM (ablation) | QD-DETR · TACoS | R1@0.5: 38.17 -> 39.37 | Improved |
| add BPM (ablation) | QD-DETR · TACoS | R1@0.5: 38.17 -> 40.04 | Improved |
| integrate query expansion and boundary modeling into CG-DETR | CG-DETR · Charades-STA | R1@0.5: 57.71 -> 58.87 | Improved |
| integrate query expansion and boundary modeling into Eatr | Eatr · TACoS | R1@0.5: 31.69 -> 34.12 | Improved |
| swap LLM for query expansion (Qwen instead of LLaMa3-8B) (ablation) | QD-DETR+Ours · Charades-STA | R1@0.5: 59.71 -> 59.51 | No clear effect |
| randomly swap start and end expanded queries (noisy queries) (ablation) | QD-DETR+Ours · Charades-STA | R1@0.5: 59.71 -> 59.23 | No clear effect |
| let the LLM freely describe objects (Prompt-Object) (ablation) | QD-DETR+Ours · TACoS | R1@0.5 and R1@0.7: 40.83 -> 39.42; 24.04 -> 21.92 | Hurt |
| limit sentences to 10 words (Prompt-Concise) (ablation) | QD-DETR+Ours · TACoS | R1@0.5: 40.83 -> 40.61 | No clear effect |
| use distance plus similarity with expanded query instead of Gaussian smoothing (ablation) | QD-DETR+Ours · TACoS | R1@0.5: 39.07 -> 40.83 | Improved |
Closest papers by meaning
- Boundary-Aware Temporal Dynamic Pseudo-Supervision Pairs Generation for Zero-Shot Natural Language Video Localization (AAAI 2025 · similarity 0.90)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · similarity 0.88)
- Weakly Supervised Temporal Action Localization via Dual-Prior Collaborative Learning Guided by Multimodal Large Language Models (CVPR 2025 · similarity 0.87)
- VideoGEM: Training-free Action Grounding in Videos (CVPR 2025 · similarity 0.87)
- Omnipotent Distillation with LLMs for Weakly-Supervised Natural Language Video Localization: When Divergence Meets Consistency (AAAI 2024 · similarity 0.87)
- Masked Diffusion Vision-Language Models for Temporal Action Localization (NeurIPS 2026 · 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.
