Research map / Video / Video-language understanding
Parallelized Autoregressive Decoding for Omni-Modal Dense Video Captioning
Wenzheng Zeng, Siyi Jiao, Chen Gao, Hwee Tou Ng, Mike Zheng Shou
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 latent global planning tokens (ablation) | Video-SALMONN 2+ 3B · ChronusAV DVC | F1 / Sim: F1 32.7 -> 61.5; Sim 17.6 -> 38.4 | Improved |
| use textual planning instead of none (ablation) | baseline · ChronusAV DVC | F1 / Sim: F1 32.7 -> 37.4; Sim 17.6 -> 19.4 | Improved |
| latent instead of textual planning (ablation) | PadCaptioner · ChronusAV DVC | F1: 37.4 -> 61.5 | Improved |
| add event-factorized parallel decoding (ablation) | PadCaptioner with planning · ChronusAV DVC | T/token: 22.9 -> 13.8 ms | Improved |
| add parallel decoding (ablation) | PadCaptioner with planning · ChronusAV DVC | F1 / Sim: F1 61.5 -> 61.8; Sim 38.4 -> 38.8 | No clear effect |
| use factorized attention rather than standard causal attention (ablation) | parallel decoding · ChronusAV DVC | F1: 38.7 -> 61.8 | Improved |
| give each branch access to all global event tokens (ablation) | factorized attention · ChronusAV DVC | F1: 59.8 -> 61.8 | Improved |
| add audio-visual semantic aggregation into event tokens (ablation) | global event tokens · ChronusAV DVC | F1: 47.7 -> 61.8 | Improved |
| attention-guided instead of mean-pooling aggregation (ablation) | global event tokens · ChronusAV DVC | F1: 58.5 -> 61.8 | Improved |
| attention-guided instead of learned scoring head (ablation) | global event tokens · ChronusAV DVC | F1: 61.4 -> 61.8 | No clear effect |
| apply method to LongVALE-LLM base model | LongVALE-LLM · ChronusAV DVC (12K) | F1 / Sim: F1 24.5 -> 36.5; Sim 21.0 -> 25.8 | Improved |
| PadCaptioner vs ChronusOmni 7B | dense video captioning · LongVALE DVC | T/video: 16162.1 -> 4283.8 ms | Improved |
Closest papers by meaning
- Event-Equalized Dense Video Captioning (CVPR 2025 · similarity 0.89)
- LongVALE: Vision-Audio-Language-Event Benchmark Towards Time-Aware Omni-Modal Perception of Long Videos (CVPR 2025 · similarity 0.89)
- Streaming Dense Video Captioning (CVPR 2024 · similarity 0.88)
- FastVID: Dynamic Density Pruning for Fast Video Large Language Models (NeurIPS 2025 · similarity 0.88)
- DIBS: Enhancing Dense Video Captioning with Unlabeled Videos via Pseudo Boundary Enrichment and Online Refinement (CVPR 2024 · similarity 0.88)
- OmniViD: A Generative Framework for Universal Video Understanding (CVPR 2024 · similarity 0.88)
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.
