Research map / Video / Video-language understanding
video-SALMONN S: Memory-Enhanced Streaming Audio-Visual LLM
Guangzhi Sun, Yixuan Li, Xiaodong Wu, Yudong Yang, Wei Li, Zejun MA, Chao Zhang
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 |
|---|---|---|---|
| replace similarity merging with TTT_MEM memory + discarding | streaming video LLM · ELViM | accuracy (%): 38.5 -> 43.9 | Improved |
| add long-span prediction loss to TTT-video (ablation) | TTT · ELViM / LVBench | accuracy (%): 42.3 -> 43.6 (ELViM) | Improved |
| add TTT-video over Mamba-2 as memory (ablation) | similarity discarding · ELViM | accuracy (%): 39.6 -> 42.3 | Improved |
| use K-means clustering instead of discarding/merging (ablation) | token downsampling · ELViM | accuracy (%): 36.8 -> 33.8 | Hurt |
| add stage 2 scale-up training (ablation) | TTT_MEM · long video benchmarks | accuracy (%): 43.6 -> 43.9 (ELViM) | Improved |
| replace AdaReTaKe with modality-aware memory reading (ablation) | video-SALMONN S · VideoEvalPro | accuracy (%): 56.5 -> 58.9 | Improved |
| add TTT_MEM to all 4 deepstack streams (ablation) | video-SALMONN S · LVBench | accuracy (%): 55.4 -> 54.7 | Hurt |
| increase memory tokens from 16k to 128k (ablation) | video-SALMONN S (w reading) · ELViM | accuracy (%): 46.7 -> 54.9 | Improved |
| increase memory reading at 128k memory (ablation) | video-SALMONN S · Video-MME | accuracy (%): 77.4 -> 77.3 | No clear effect |
| change long-span prediction span T from 2 to 1 or 32 (ablation) | TTT_MEM · Video-MME long | accuracy (%): 71.3 -> 70.6 (T=1) | Hurt |
| use video-SALMONN S instead of non-streaming Qwen3-VL on VideoEvalPro | Qwen3-VL 8B · VideoEvalPro | accuracy (%): 54.9 -> 58.9 | Improved |
| use PEMF or similarity merging versus non-streaming baseline | streaming baselines · VideoEvalPro | accuracy (%): 54.9 -> 51.8 | Hurt |
Closest papers by meaning
- video-SALMONN: Speech-Enhanced Audio-Visual Large Language Models (ICML 2024 · similarity 0.90)
- FreshMem: Brain-Inspired Frequency-Space Hybrid Memory for Streaming Video Understanding (NeurIPS 2026 · similarity 0.89)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.88)
- video-SALMONN-o1: Reasoning-enhanced Audio-visual Large Language Model (ICML 2025 · similarity 0.88)
- StreamOV: Streaming Omni-Video Understanding via Evidence-Guided Memory and Response Triggering (NeurIPS 2026 · similarity 0.88)
- StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025 · 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.
