AVTrace: Diagnosing Audio-Visual Temporal Reasoning in Omni Models
arXiv:2609.19991v1 Announce Type: cross Abstract: Omni models can describe video content, but can they locate events in time, preserve event order, and judge audio-visual synchronization? We introduce AVTrace (Audio-Visual Temporal Reasoning…
Read the full story at arXiv cs.AI ↗
ImpactNotable 31/100
Why it mattersRule-based estimate: event keywords (+4); trust 6/10.
RegionsGlobal
Published1 d ago (Fri, 18 Sep 2026 04:00:00 GMT)
RetrievedFri, 18 Sep 2026 08:00:48 GMT via rss
ClassifiedFri, 18 Sep 2026 08:01:04 GMT by heuristic
AuthorLongyin Zhang, Parth Sakhare Mahendra, Chengwei Wei, Ning Zhang, Lim Ming Chong, Sirui He, Ai Ti Aw