Research map / Video / Video-language understanding
Context-Aware Biaffine Localizing Network for Temporal Sentence Grounding
Daizong Liu, Xiaoye Qu, Jianfeng Dong, Pan Zhou, Yu Cheng, Wei Wei, Zichuan Xu, Yulai Xie
Check 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 biaffine scoring of all start-end pairs instead of boundary regression (ablation) | TSG baseline · ActivityNet Captions | R@1, IoU=0.7: 17.02 -> 22.74 | Improved |
| Use biaffine scoring of all start-end pairs instead of proposal matching (ablation) | TSG baseline · ActivityNet Captions | R@1, IoU=0.7: 20.10 -> 22.74 | Improved |
| Add MMSA to biaffine baseline (ablation) | CBLN · ActivityNet Captions | R@1, IoU=0.7: 22.74 -> 24.22 | Improved |
| Add MCBL local-global context aggregation to biaffine baseline (ablation) | CBLN · ActivityNet Captions | R@1, IoU=0.7: 22.74 -> 26.48 | Improved |
| Use local-frame concatenation instead of mean-pooling or max-pooling (ablation) | CBLN MCBL · ActivityNet Captions | R@1, IoU=0.7: 25.09 / 25.47 -> 27.60 | Improved |
| Use global max-pooling rather than sampling or mean-pooling (ablation) | CBLN MCBL · ActivityNet Captions | R@1, IoU=0.7: 24.88 / 26.02 -> 27.60 | Improved |
| Use multi-scale local windows {1,3,5} rather than a single local scale {1} (ablation) | CBLN MCBL · ActivityNet Captions | R@1, IoU=0.7: 25.45 -> 27.60 | Improved |
| Add local scale 7 to {1,3,5} (ablation) | CBLN MCBL · ActivityNet Captions | R@1, IoU=0.7: 27.60 -> 27.38 | Hurt |
| Add global scale 8 to {1,2,4} (ablation) | CBLN MCBL · ActivityNet Captions | R@1, IoU=0.7: 27.60 -> 27.72 | Improved |
| Add non-local blocks to reweight local and global contexts (ablation) | CBLN MCBL · ActivityNet Captions | R@1, IoU=0.7: 25.05 -> 27.60 | Improved |
| Use non-local blocks instead of linear layers for context interaction (ablation) | CBLN MCBL · ActivityNet Captions | R@1, IoU=0.7: 25.93 -> 27.60 | Improved |
| Use a final linear layer instead of pooling to combine local/global contexts (ablation) | CBLN MCBL · ActivityNet Captions | R@1, IoU=0.7: 26.86 -> 27.60 | Improved |
Closest papers by meaning
- Boundary-Aware Temporal Dynamic Pseudo-Supervision Pairs Generation for Zero-Shot Natural Language Video Localization (AAAI 2025 · similarity 0.88)
- Temporal Sentence Grounding with Relevance Feedback in Videos (NeurIPS 2024 · similarity 0.88)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.88)
- Multi-Pair Temporal Sentence Grounding via Multi-Thread Knowledge Transfer Network (AAAI 2025 · similarity 0.88)
- Unsupervised Domain Adaptative Temporal Sentence Localization with Mutual Information Maximization (AAAI 2024 · similarity 0.88)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (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.
