Research map / Video / Video-language understanding
DASH: Dynamic Audio-Driven Semantic Chunking for Efficient Omnimodal Token Compression
Bingzhou Li, Tao Huang
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 |
|---|---|---|---|
| use DASH at 25% retention instead of OmniZip at 35% | Qwen2.5-Omni-7B · AVUT avg | accuracy: 60.6 -> 60.9 | Improved |
| use DASH at 25% retention instead of OmniZip at 35% | Qwen2.5-Omni-7B · WorldSense avg | accuracy: 44.7 -> 44.9 | Improved |
| use DASH at 25% retention instead of OmniZip at 35% | Qwen2.5-Omni-3B · WorldSense avg | accuracy: 44.1 -> 44.6 | Improved |
| use DASH at 25% retention instead of OmniZip at 35% | Qwen2.5-Omni-3B · AVUT avg | accuracy: 58.7 -> 58.8 | No clear effect |
| use DASH at 25% retention instead of OmniZip at 35% | Qwen2.5-Omni-3B · VideoMME w/o subtitles | accuracy: 61.9 -> 61.7 | Hurt |
| use DASH at 35% retention instead of OmniZip at 35% | Qwen2.5-Omni-7B · AVUT avg | accuracy: 60.6 -> 61.5 | Improved |
| add tri-signal fusion to static grouping (ablation) | OmniZip-style static grouping (3B, 25% retention) · WorldSense avg | accuracy: 44.1 -> 44.4 | Improved |
| add dynamic semantic chunking with attention-only selection (ablation) | OmniZip-style static grouping (3B, 25% retention) · WorldSense avg | accuracy: 44.1 -> 44.4 | Improved |
| combine dynamic chunking and tri-signal fusion (ablation) | DASH (3B, 25% retention) · WorldSense avg | accuracy: 44.4 -> 44.6 | Improved |
| use cosine similarity instead of dot product for boundary detection (ablation) | DASH (3B, 25% retention) · WorldSense avg | accuracy: 43.6 -> 44.6 | Improved |
| use cosine boundaries instead of random boundaries (ablation) | DASH (3B, 25% retention) · WorldSense avg | accuracy: 43.8 -> 44.6 | Improved |
| use cosine instead of change rate for boundary detection (ablation) | DASH (3B, 25% retention) · WorldSense avg | accuracy: 44.0 -> 44.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- OmniZip — OmniZip: Audio-Guided Dynamic Token Compression for Fast Omnimodal Large Language Models (CVPR 2026) · 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
- OmniSelect: Dynamic Modality-Aware Token Compression for Efficient Omni-modal Large Language Models (NeurIPS 2026 · similarity 0.91)
- ALMTokenizer: A Low-bitrate and Semantic-rich Audio Codec Tokenizer for Audio Language Modeling (ICML 2025 · similarity 0.89)
- AVOC: Enhancing Hour-Level Audio-Video Understanding in Omni-Modal LLMs via Retrieval-Inspired Token Compression (NeurIPS 2026 · similarity 0.89)
- WavTokenizer: an Efficient Acoustic Discrete Codec Tokenizer for Audio Language Modeling (ICLR 2025 · similarity 0.87)
- EvoComp: Learning Visual Token Compression for Multimodal Large Language Models via Semantic-Guided Evolutionary Labeling (CVPR 2026 · similarity 0.87)
- One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding (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.
