LAT-Chronicle is the first large-scale long-form audio dataset designed for temporal awareness, comprising approximately 1,200 hours of audio, including 1,000 hours in Chinese and 200 hours in English.
It provides high-quality temporal annotations across diverse real-world scenarios, enabling research on long-form audio temporal awareness.
LAT-Chronicle covers three core tasks:
Table 1: Overview of LAT-Chronicle.
Figure 1: Overview of LAT-Pipe. The pipeline constructs LAT-Chronicle through diverse audio sourcing, multi-track atomic annotation, task-oriented label generation, and human-in-the-loop quality control.
LAT-Pipe enables scalable and high-quality temporal annotation for long-form audio.
Table 2: Audio scenario taxonomy in LAT-Pipe.
LAT-Chronicle provides balanced coverage across durations and diverse real-world scenarios.
Figure 2: Duration and scenario distributions of LAT-Chronicle across Chinese and English.
Annotations cover full temporal ranges with consistent event density across tasks.
Table 3: Temporal annotation statistics of LAT-Chronicle. Avg. Evt. = average number of events; Avg. Evt. Dur = average event duration (seconds).
LAT-Chronicle exhibits high temporal density, with DAC containing a large number of densely annotated events per audio. For TAG and TAC, the target intervals are evenly distributed across the beginning, middle, and end of each audio sample.
LAT-Chronicle is organized into two types of files: metadata files and task files.
./meta/chronicle-CN-meta.jsonl./meta/chronicle-EN-meta.jsonlThese files provide metadata for each audio sample, including:
id: unique identifierurl: source link for downloading the audiotitle: original audio titleduration: duration in secondsDense Audio Captioning (DAC)
./task/chronicle-CN-DAC.jsonl./task/chronicle-EN-DAC.jsonlTemporal Audio Grounding (TAG)
./task/chronicle-CN-TAG.jsonl./task/chronicle-EN-TAG.jsonlTargeted Audio Captioning (TAC)
./task/chronicle-CN-TAC.jsonl./task/chronicle-EN-TAC.jsonlEach task file contains benchmark instances in a unified format. The audios field references the corresponding audio sample using the id from metadata files.
For detailed evaluation protocols and metrics, please refer to the official repository:
๐ https://github.com/alanshaoTT/LAT-Audio-Repo
If you find this work useful, please cite:
@article{shao2026lataudio,
title={Listening with Time: Precise Temporal Awareness for Long-Form Audio Understanding},
author={Shao, Mingchen and Su, Hang and Tian, Wenjie and Mu, Bingshen and Lin, Zhennan and Fan, Lichun and Luo, Zhenbo and Luan, Jian and Xie, Lei},
journal={arXiv preprint arXiv:2604.22245},
year={2026}
}
For questions, feedback, or collaboration inquiries, please contact:
14 commits
LAT-Chronicle is the first large-scale long-form audio dataset designed for temporal awareness, comprising approximately 1,200 hours of audio, including 1,000 hours in Chinese and 200 hours in English.
It provides high-quality temporal annotations across diverse real-world scenarios, enabling research on long-form audio temporal awareness.
LAT-Chronicle covers three core tasks:
Table 1: Overview of LAT-Chronicle.
Figure 1: Overview of LAT-Pipe. The pipeline constructs LAT-Chronicle through diverse audio sourcing, multi-track atomic annotation, task-oriented label generation, and human-in-the-loop quality control.
LAT-Pipe enables scalable and high-quality temporal annotation for long-form audio.
Table 2: Audio scenario taxonomy in LAT-Pipe.
LAT-Chronicle provides balanced coverage across durations and diverse real-world scenarios.
Figure 2: Duration and scenario distributions of LAT-Chronicle across Chinese and English.
Annotations cover full temporal ranges with consistent event density across tasks.
Table 3: Temporal annotation statistics of LAT-Chronicle. Avg. Evt. = average number of events; Avg. Evt. Dur = average event duration (seconds).
LAT-Chronicle exhibits high temporal density, with DAC containing a large number of densely annotated events per audio. For TAG and TAC, the target intervals are evenly distributed across the beginning, middle, and end of each audio sample.
LAT-Chronicle is organized into two types of files: metadata files and task files.
./meta/chronicle-CN-meta.jsonl./meta/chronicle-EN-meta.jsonlThese files provide metadata for each audio sample, including:
id: unique identifierurl: source link for downloading the audiotitle: original audio titleduration: duration in secondsDense Audio Captioning (DAC)
./task/chronicle-CN-DAC.jsonl./task/chronicle-EN-DAC.jsonlTemporal Audio Grounding (TAG)
./task/chronicle-CN-TAG.jsonl./task/chronicle-EN-TAG.jsonlTargeted Audio Captioning (TAC)
./task/chronicle-CN-TAC.jsonl./task/chronicle-EN-TAC.jsonlEach task file contains benchmark instances in a unified format. The audios field references the corresponding audio sample using the id from metadata files.
For detailed evaluation protocols and metrics, please refer to the official repository:
๐ https://github.com/alanshaoTT/LAT-Audio-Repo
If you find this work useful, please cite:
@article{shao2026lataudio,
title={Listening with Time: Precise Temporal Awareness for Long-Form Audio Understanding},
author={Shao, Mingchen and Su, Hang and Tian, Wenjie and Mu, Bingshen and Lin, Zhennan and Fan, Lichun and Luo, Zhenbo and Luan, Jian and Xie, Lei},
journal={arXiv preprint arXiv:2604.22245},
year={2026}
}
For questions, feedback, or collaboration inquiries, please contact:
14 commits