OmniSeek: Native Tool Integration for Multi-turn Audio-Visual Reasoning

Benchmark Model Rank Results
audio-visual-question-answering-on-dailyomniOmniSeek (ours)–Accuracy: 80.0
audio-visual-question-answering-on-futureomniOmniSeek (ours)–Accuracy: 58.3
audio-visual-question-answering-on-lvomnibenchOmniSeek (ours)–Accuracy: 44.2
audio-visual-question-answering-on-omnivideobenchOmniSeek (ours)–Accuracy: 47.7
audio-visual-question-answering-on-worldsenseOmniSeek (ours)–Accuracy: 62.4
video-question-answering-on-lvbenchOmniSeek (ours)–Accuracy: 51.4
video-question-answering-on-mlvuOmniSeek (ours)–Accuracy: 77.1
video-question-answering-on-videoholmesOmniSeek (ours)–Accuracy: 74.6