Time Markers Ground Audio Answers Across Two Hours
GigaChat Audio interleaves timestamp anchors with continuous audio tokens, reaching 65.2 mIoU on 20–40 minute grounding when trained with 7-second anchors.
Jul 30, 20265 min2607.10387
2 articles on SOTA Papers
GigaChat Audio interleaves timestamp anchors with continuous audio tokens, reaching 65.2 mIoU on 20–40 minute grounding when trained with 7-second anchors.
A Conformer encoder trained on 2M hours uses cluster-level balancing and domain-aware fine-tuning to reduce head-language dominance.
