AVTrace: Diagnosing Audio-Visual Temporal Reasoning in Omni Models

AVTrace is a diagnostic suite for evaluating the temporal reasoning capabilities of omni models. Five open omni models were tested using AVTrace, revealing weaknesses in synchronization verification, chain parsing, and event-conditioned grounding and comprehension. This research provides insights into the limitations of current omni models and the importance of temporal post-training.

RSS Score 0 9/18/2026, 4:00:00 AM Original Source
Save an API key to vote.