aiqbench analysis
This paper introduces Logbook, a new benchmark designed to evaluate AI systems on hour-scale audio understanding, moving beyond the limitations of short, pre-segmented clips. The study compares 52 different architectures and finds that while end-to-end models often outperform cascaded approaches, current systems still struggle with over-segmentation and fall short of human reference levels even when fine-tuned.
Logbook: Extremely Long-form Audio Event Understanding
arXiv:2610.07338v1 Announce Type: cross Abstract: Audio benchmarks are built around short, pre-segmented clips, limiting model design to brief inputs or fixed vocabularies. To close this gap, we introduce Logbook, a benchmark for hour-scale audio understanding, with recordings ranging from ten minutes to six days. Given a continuous audio recording and an event label vocabulary, a system must predict a gap-free segmentation with an event label and a description per segment. We compare 52 systems, end-to-end and cascaded, and ablate fine-tuning, context length, and reasoning budget. We find the task tractable, though the best systems remain below the human reference. Also, over-segmentation is pervasive, and fine-tuning partially mitigates it. Finally, end-to-end are often better than cascaded systems, but degrades with longer context.
Abstract:Audio benchmarks are built around short, pre-segmented clips, limiting model design to brief inputs or fixed vocabularies. To close this gap, we introduce Logbook, a benchmark for hour-scale audio understanding, with recordings ranging from ten minutes to six days. Given a continuous audio recording and an event label vocabulary, a system must predict a gap-free segmentation with an event label and a description per segment. We compare 52 systems, end-to-end and cascaded, and ablate fine-tuning, context length, and reasoning budget. We find the task tractable, though the best systems remain below the human reference. Also, over-segmentation is pervasive, and fine-tuning partially mitigates it. Finally, end-to-end are often better than cascaded systems, but degrades with longer context.
Submitted to ICASSP 2027. Source code available at this https URL
Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Sound (cs.SD)
arXiv:2610.07338 [eess.AS]
(or arXiv:2610.07338v1 [eess.AS] for this version)
https://doi.org/10.48550/arXiv.2610.07338 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 235 words
