AI Paper Atlas

Research map / Video / Video-language understanding

Video-Oasis: Rethinking Evaluation of Video Understanding

Geuntaek Lim, Minho Shim, Sungjune Park, Jaeyun Lee, Inwoong Lee, Taeoh Kim, Dongyoon Wee, Yukyung Choi

ECCV 2026 · Video › Video-language understanding · approach: Long video understanding · 3 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 AKS temporal grounding (16 frames) (ablation)Eagle2.5 · Video-Oasis distilled setaccuracy: 31.5 -> 32.9Improved
add AKS temporal grounding (16 frames) (ablation)Qwen3-VL-Instruct · Video-Oasis distilled setaccuracy: 27.8 -> 30.1Improved
add oracle ground-truth temporal grounding (ablation)Eagle2.5 · Video-Oasis distilled samplesaccuracy: 35.0 -> 50.8Improved
add oracle ground-truth temporal grounding (ablation)Eagle2.5 · shortcut samplesaccuracy: 78.0 -> 80.8No clear effect
use adaptive thinking (VideoAuto-R1) instead of always-on thinking (ablation)Qwen3-VL-8B · Video-Oasis distilled setaccuracy: 34.6 -> 36.8Improved
switch from instruct to thinking mode (ablation)Qwen3-VL-8B · Video-Oasis distilled setaccuracy: 33.8 -> 34.6Improved
oracle choice between instruct and thinking per question (ablation)Qwen3-VL-8B · Video-Oasis distilled setaccuracy: 34.6 -> 46.2Improved
long-context SFT (Eagle2.5) over baseQwen2.5-VL · Video-Oasis distilled setaccuracy: 29.2 -> 34.5Improved
RLVR with QA-only reward (Video-R1)Qwen2.5-VL · Video-Oasis distilled setaccuracy: 29.2 -> 26.3Hurt
RLVR with QA plus grounding reward (VideoAuto-R1)Qwen2.5-VL · Video-Oasis distilled setaccuracy: 29.2 -> 32.7Improved
add grounding reward in RLVRQwen2.5-VL · Global Narrativeaccuracy: 21.2 -> 28.6Improved
filter shortcut samples with Video-OasisQwen2.5-VL (7B) · 14 benchmarksaccuracy: 51.2 -> 29.2Hurt

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