Research map / Video / Video-language understanding
HieraMamba: Video Temporal Grounding via Hierarchical Anchor-Mamba Pooling
Joungbin An, Kristen Grauman
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 bidirectional scan (Hydra) (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 23.29 -> 25.66 | Improved |
| add anchor interleaving (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 24.40 -> 25.66 | Improved |
| add local encoding (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 24.63 -> 25.66 | Improved |
| add learnable gates (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 24.80 -> 25.66 | Improved |
| add ACC loss (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 24.68 -> 25.22 | Improved |
| add SPC loss (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 24.68 -> 25.23 | Improved |
| add both ACC and SPC (ablation) | HieraMamba · TACoS | Avg. recall: 65.66 -> 66.65 | Improved |
| add SPC loss alone (ablation) | HieraMamba · MAD-v2 | Avg. recall: 17.29 -> 17.35 | No clear effect |
| use unpooled SPC (every token positive) instead of pooled prototype (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 24.68 -> 24.05 | Hurt |
| learned anchor pooling (attention/gated) instead of mean pooling (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 24.68 -> 24.46 / 24.37 | Hurt |
| increase AMP depth from 8 to 10 layers (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 25.66 -> 24.84 | Hurt |
| reduce AMP depth to 2 layers (ablation) | HieraMamba · Ego4D-NLQ | Avg. recall: 25.66 -> 21.80 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DeCafNet — DeCafNet: Delegate and Conquer for Efficient Temporal Grounding in Long Videos (CVPR 2025) · better on 3 of 3 shared results
- RGNet — RGNet: A Unified Clip Retrieval and Grounding Network for Long Videos (ECCV 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · similarity 0.90)
- DART: Difficulty-Adaptive Routing for Zero-Shot Video Temporal Grounding (ECCV 2026 · similarity 0.90)
- Scanning Only Once: An End-to-end Framework for Fast Temporal Grounding in Long Videos (ICCV 2023 · similarity 0.90)
- HERO: Hierarchical Embedding-Refinement for Open-Vocabulary Temporal Sentence Grounding in Videos (CVPR 2026 · similarity 0.89)
- ReVisionLLM: Recursive Vision-Language Model for Temporal Grounding in Hour-Long Videos (CVPR 2025 · similarity 0.88)
- Omnipotent Distillation with LLMs for Weakly-Supervised Natural Language Video Localization: When Divergence Meets Consistency (AAAI 2024 · 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.
