Research map / Video / Video-language understanding
Selective Query-Guided Debiasing for Video Corpus Moment Retrieval
Sunjae Yoon, Ji Woo Hong, Eunseop Yoon, Dahyun Kim, Junyeong Kim, Hee Suk Yoon, Chang D. Yoo
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 biased moment retrieval (BMR) with selective debiasing (ablation) | SQuiDNet · VCMR TVR val | R@1 (tIoU=0.7): 4.62 -> 8.52 | Improved |
| add conditional moment prediction (CMP) (ablation) | SQuiDNet · VCMR TVR val | R@1 (tIoU=0.7): 8.17 -> 8.52 | Improved |
| use selective decision rule instead of treating all BMR predictions as negative (ablation) | SQuiDNet · VCMR TVR val | R@1 (tIoU=0.7): 6.41 -> 8.52 | Improved |
| use selective decision rule instead of treating all BMR predictions as positive (ablation) | SQuiDNet · VCMR TVR val | R@1 (tIoU=0.7): 6.72 -> 8.52 | Improved |
| use co-occurrence table decision rule instead of all-negative (ablation) | SQuiDNet · VCMR TVR val | R@1 (tIoU=0.7): 6.41 -> 7.91 | Improved |
| use learnable confounder instead of co-occurrence table (ablation) | SQuiDNet · VCMR TVR val | R@1 (tIoU=0.7): 7.91 -> 8.52 | Improved |
| remove subtitles and use plain video features instead of subtitle-matched features (ablation) | SQuiDNet · VCMR DiDeMo | R@1/R@10/R@100 (tIoU=0.7): -0.36%/-0.74%/-1.82% | Hurt |
| add BMR to a HERO baseline NMR | HERO · VCMR TVR test-public | R@1 (tIoU=0.7): 6.21 -> 8.34 | Improved |
| add BMR to the naive model, with SQuiD debiasing | SQuiDNet (N) · VCMR TVR test-public | R@1 (tIoU=0.7): 4.09 -> 10.09 | Improved |
| add BMR to the naive model, with SQuiD debiasing | SQuiDNet (N) · VCMR ActivityNet | R@1 (tIoU=0.7): 1.62 -> 4.43 | Improved |
| add BMR to the naive model, with SQuiD debiasing | SQuiDNet (N) · VCMR DiDeMo +ASR | R@1 (tIoU=0.7): 1.73 -> 3.52 | Improved |
Closest papers by meaning
- Fewer Steps, Better Performance: Efficient Cross-Modal Clip Trimming for Video Moment Retrieval Using Language (AAAI 2024 · similarity 0.89)
- Reverse Distribution Based Video Moment Retrieval for Effective Bias Elimination (AAAI 2025 · similarity 0.89)
- Generative Video Diffusion for Unseen Novel Semantic Video Moment Retrieval (AAAI 2025 · similarity 0.88)
- VERIFIED: A Video Corpus Moment Retrieval Benchmark for Fine-Grained Video Understanding (NeurIPS 2024 · similarity 0.88)
- MomentSeeker: A Task-Oriented Benchmark For Long-Video Moment Retrieval (NeurIPS 2025 · similarity 0.87)
- Unbiasing through Textual Descriptions: Mitigating Representation Bias in Video Benchmarks (CVPR 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.
