Research map / Video / Video-language understanding
HAS: Highlight-guided Attention Steering for Multimodal LLM Video Summarization
Rui Chu, Yingjie Lao
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 |
|---|---|---|---|
| apply HAS attention steering to the video MLLM at inference time | V2Xum-LLaMA-level summarizers · TVSum | Kendall tau: 0.222 -> 0.224 | Improved |
| HAS vs best baseline on SumMe Spearman | HAS · SumMe | Spearman rho: 0.378 -> 0.335 | Hurt |
| HAS vs V2Xum-LLaMA on VideoXum cross-modal metrics | HAS · VideoXum | CIDEr / F1 / FCLIP: CIDEr 26.9 -> 28.0; F1 31.6 -> 32.0; FCLIP 0.957 -> 0.963 | Improved |
| HAS vs full-context zero-shot MLLM backbones on grounded summarization | open-source video MLLMs · VISTA (zero-shot) | FactVC: 42.07 -> 50.11 (averaged HAS vs mPLUG-Owl3) | Improved |
| HAS (training-free) vs fully fine-tuned models on VISTA | HAS · VISTA | FactVC: 75.41 -> 50.11 | Hurt |
| soft attention steering instead of hard frame selection (ablation) | HAS · VISTA coverage vs budget T | ROUGE-Lsum / fact recall | Improved |
| zero-shot transfer from SumMe to MR.HiSum | HAS · MR.HiSum | Kendall tau / Spearman rho | 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.
- CSTA — CSTA: CNN-based Spatiotemporal Attention for Video Summarization (CVPR 2024) · better on 1 of 1 shared result
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
- Video Summarization with Large Language Models (CVPR 2025 · similarity 0.92)
- V2Xum-LLM: Cross-Modal Video Summarization with Temporal Prompt Instruction Tuning (AAAI 2025 · similarity 0.90)
- SF2T: Self-supervised Fragment Finetuning of Video-LLMs for Fine-Grained Understanding (CVPR 2025 · similarity 0.89)
- MMSum: A Dataset for Multimodal Summarization and Thumbnail Generation of Videos (CVPR 2024 · similarity 0.89)
- SlowFocus: Enhancing Fine-grained Temporal Understanding in Video LLM (NeurIPS 2024 · similarity 0.89)
- Scaling Up Video Summarization Pretraining with Large Language Models (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.
