LVOmniBench
LVOmniBench: Pioneering Long Audio-Video Understanding Evaluation for Omnimodal LLMs
Keda Tao, Yuhua Zheng, Jia Xu, Wenjie Du, Kele Shao, Hesong Wang, et al. · 2026-03-19
- Modality
- Video & streaming
- Data
- Real data
- Task · ground truth
- QA · Ans.
- Scope
- Broader evaluation with memory demands
What it tests
- CECross-modal evidenceDirectly evaluated
- LELong-range evidenceDirectly evaluated
- SCState changesPartially evaluated
- CFConflicting memoriesNot evaluated
- UQUnanswerable queriesNot evaluated
- MAMemory-grounded actionNot evaluated
- TLTemporal localizationPartially evaluated
- TRTemporal relationsDirectly evaluated
- MEMulti-event integrationDirectly evaluated
Survey notes
Full-recording audio-video QA targets perception, understanding, inference and logical reasoning. Long inputs and temporal reasoning do not by themselves define an agent-memory evaluation.