Metraipolis › archive › dc_66adde07bfe0

LongCap: Hour-Scale Timestamped Audio Captioning via Chunk-wise Processing — arXiv:2611.11125

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.11125
License (as archived): CC0-1.0
Archived: 2026-11-10T18:44:35.597Z
SHA-256 of the archived text: fc37bd15ece6353c8e11564d996ddc99e70662360b65284d15301bb7f51d227c
Stable URL of this page: https://metraipolis.com/documento/dc_66adde07bfe0

How to cite

LongCap: Hour-Scale Timestamped Audio Captioning via Chunk-wise Processing — arXiv:2611.11125 — arXiv (en), https://arxiv.org/abs/2611.11125 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_66adde07bfe0 (Metraipolis archive dc_66adde07bfe0, sha256 fc37bd15ece6353c8e11564d996ddc99e70662360b65284d15301bb7f51d227c)

BibTeX

@misc{dc66adde07bfe0,
  title = {LongCap: Hour-Scale Timestamped Audio Captioning via Chunk-wise Processing — arXiv:2611.11125},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.11125},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_66adde07bfe0; sha256 fc37bd15ece6353c8e11564d996ddc99e70662360b65284d15301bb7f51d227c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.11125. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Describing hours of audio requires accurate event timing and sustained descriptive detail throughout the recording. We present LongCap, a framework for timestamped audio captioning (TAC), with the human-annotated LongCapBench and three diarization-inspired metrics: Timeline Error Rate (TER), Granularity Error Rate (GER), and Caption Similarity (tcpCapSim). LongCap interleaves audio chunks with structured event captions, using relative timestamps and retained context to support length extrapolation within a 1M-token inference context window. Event-level reannotation restores caption detail, while systematic studies examine model scaling, dependency supervision, and self-distillation. LongCap-Diar achieves strong benchmark results and extrapolates to 16 hours of concatenated audio despite training on recordings no longer than 40 minutes. Our TAC models are competitive with Qwen3.5-Omni-Plus on short clips, outperform evaluated baselines in caption similarity and timeline accuracy in segment-level comparisons on LongCapBench-TAC-Long, and caption full recordings approaching two hours. On LongCapBench-TAC, 9B and 35B-A3B offer favorable size-performance tradeoffs, while 122B-A10B substantially improves event-dependency recovery. These findings support long-horizon general audio understanding and provide a foundation for always-on voice agents.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Sound (cs.SD).
Pagina del lavoro: https://arxiv.org/abs/2611.11125 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-10T18:37:52.435Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).