AI Paper Atlas

Research map / Video / Video-language understanding

Towards Sparse Video Understanding and Reasoning

Chenwei Xu, Zhen Ye, Shang Wu, Weijian Li, Zihan Wang, Zhuofan Xia, Lie Lu, Pranav Maneriker, Fan Du, Manling Li, Han Liu

CVPR 2026 · Video › Video-language understanding · approach: Video question answering · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
add state carryover (summary-as-state) (ablation)ReViSe · VideoEspressoaccuracy (%): 23.14 -> 41.48Improved
add structured P/O/H/U/R summary fields (ablation)ReViSe · VideoEspressoaccuracy (%): 24.27 -> 41.48Improved
add both state carryover and structured fields (ablation)ReViSe · VideoEspressoaccuracy (%): 20.24 -> 41.48Improved
allow 4 turns instead of 1 (ablation)ReViSe · VideoEspressoaccuracy (%): 38.3 -> 42.1Improved
allow 4 turns instead of 1 (ablation)ReViSe · VideoEspressoframes used: 4.60 -> 2.89Improved
balanced rounds-frames allocation (5x4) instead of skewed 10x2 (ablation)ReViSe · VideoEspressoaccuracy (%): 63.3 -> 66.9Improved
increase max rounds from 5 to 7 (4 frames each) (ablation)ReViSe · VideoEspressoaccuracy (%): 66.9 -> 67.4Mixed
reinforcement fine-tuning (EAGER GRPO) instead of plug-and-playReViSe Qwen2.5-VL-3B · NExT-QAaccuracy (%): 31.7 -> 51.3Improved
reinforcement fine-tuning instead of supervised format fine-tuningReViSe Qwen2.5-VL-3B · VideoEspressoaccuracy (%): 21.3 -> 27.8Improved
drop confidence-gain term (lambda1 = 0) (ablation)EAGER reward · NExT-QAaccuracy (%): 44.4 -> 46.4No clear effect
add correct-and-early-stop term (lambda3 = 0.5) (ablation)EAGER reward · NExT-QAaccuracy (%): 39.9 -> 44.4Improved
add ReViSe to GPT-4oGPT-4o · VideoEspressoaccuracy (%): 26.4 -> 48.9Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 3 on most 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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy