{"ID":6536475,"CreatedAt":"2026-07-14T01:21:01.169441415Z","UpdatedAt":"2026-07-14T16:11:02.601666889Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.10387","arxiv_id":"2607.10387","title":"GigaChat Audio: Time-aware Large Audio Language Model","abstract":"Temporal grounding in long recordings remains challenging for audio-conditioned LLMs. We present a time-aware audio LLM that answers questions with explicit timestamps over up to 120 minutes of input. Our approach interleaves periodic time markers with continuous audio tokens using large-scale synthetic supervision from a cascaded pipeline. Our model achieves strong temporal-grounding accuracy on short and long benchmarks and supports time-anchored fragment descriptions and summaries. Extensive ablations examine how time representation, marker frequency, tokenization, and duration-mixture design affect accuracy and computational cost. We release model weights and datasets to support further research on time-aware audio understanding, available at https://huggingface.co/ai-sage/GigaChat3.1-Audio-10B-A1.8B.","short_abstract":"Temporal grounding in long recordings remains challenging for audio-conditioned LLMs. We present a time-aware audio LLM that answers questions with explicit timestamps over up to 120 minutes of input. Our approach interleaves periodic time markers with continuous audio tokens using large-scale synthetic supervision fro...","url_abs":"https://arxiv.org/abs/2607.10387","url_pdf":"https://arxiv.org/pdf/2607.10387v1","authors":"[\"Aleksandr Kutsakov\",\"Mariia Sadovina\",\"Georgii Gospodinov\",\"Alexandr Maximenko\",\"Oleg Kutuzov\",\"Pavel Bogomolov\",\"Fyodor Minkin\"]","published":"2026-07-11T16:34:25Z","proceeding":"eess.AS","tasks":"[\"eess.AS\",\"cs.CL\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false}