Research map / Video / Video-language understanding
HiVid: LLM-Guided Video Saliency For Content-Aware VOD And Live Streaming
Jiahui Chen, Bo Peng, Lianchen Jia, Zeyu Zhang, Tianchi Huang, Lifeng Sun
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 LLM-guided ranking module (ablation) | HiVid · Youtube-8M saliency | PLCC: 0.611 -> 0.660 | Improved |
| add perception summary module (ablation) | HiVid · Youtube-8M saliency | PLCC: 0.632 -> 0.660 | Improved |
| add Gaussian smoothing (ablation) | HiVid · Youtube-8M saliency | PLCC: 0.619 -> 0.660 | Improved |
| add correlation loss to MSE (ablation) | HiVid-M · time series forecasting | PLCC: 0.21 -> 0.29 | Improved |
| use adaptive output length instead of constant L_out (ablation) | HiVid-M · live overall PLCC m=10 | PLCC: 0.10/0.17/0.18 -> 0.20 | Improved |
| use GPT-4o instead of open-source MLLMs (ablation) | HiVid · Youtube-8M saliency | PLCC: 0.58-0.64 -> 0.66 | Improved |
| use LLM prompting for forecasting instead of trained model | LLM forecaster · time series forecasting | RMSE: HiVid-M 0.12 -> LLM 0.27 | Hurt |
| use more complex forecasting architectures | iTransformer/TimesNet etc. · time series forecasting | PLCC: HiVid-U 0.24 vs Crossformer 0.23 | No clear effect |
| increase window length m from 2 to 10 (ablation) | HiVid · VOD overhead | total cost per video ($): 8.12 -> 1.35 | Improved |
| increase window length m (ablation) | HiVid live · live utility | forecasting utility | Hurt |
| use GPT-4o-mini with m=2 instead of GPT-4o m=10 (ablation) | HiVid · VOD saliency | PLCC: 0.660 -> 0.645; cost 1.35 -> 0.44 | Hurt |
Closest papers by meaning
- LiveStar: Live Streaming Assistant for Real-World Online Video Understanding (NeurIPS 2025 · similarity 0.86)
- VideoLLM-MoD: Efficient Video-Language Streaming with Mixture-of-Depths Vision Computation (NeurIPS 2024 · similarity 0.85)
- Recurrent Attention-based Token Selection for Efficient Streaming Video-LLMs (NeurIPS 2025 · similarity 0.85)
- VideoLLM-online: Online Video Large Language Model for Streaming Video (CVPR 2024 · similarity 0.85)
- What Should a Streaming Video Model Remember? (NeurIPS 2026 · similarity 0.85)
- Think-as-You-See: Streaming Chain-of-Thought Reasoning for Large Vision-Language Models (CVPR 2026 · similarity 0.84)
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.
