Metraipolis › archive › dc_41fa7858c44e
Acoustic-to-Text KV Compression for Full-Duplex Speech Models — arXiv:2609.31224
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.31224
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T01:39:47.560Z
- SHA-256 of the archived text
0f86576c8f0d5b9c2732a6f4033111258d5dd7bdbe8f262fa00e1e7b1e794ee8- Stable URL of this page
https://metraipolis.com/documento/dc_41fa7858c44e
How to cite
Acoustic-to-Text KV Compression for Full-Duplex Speech Models — arXiv:2609.31224 — arXiv (en), https://arxiv.org/abs/2609.31224 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_41fa7858c44e (Metraipolis archive dc_41fa7858c44e, sha256 0f86576c8f0d5b9c2732a6f4033111258d5dd7bdbe8f262fa00e1e7b1e794ee8)
BibTeX
@misc{dc41fa7858c44e,
title = {Acoustic-to-Text KV Compression for Full-Duplex Speech Models — arXiv:2609.31224},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.31224},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_41fa7858c44e; sha256 0f86576c8f0d5b9c2732a6f4033111258d5dd7bdbe8f262fa00e1e7b1e794ee8},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.31224. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Full-duplex speech language models continuously accumulate acoustic key-value (KV) states, making long-running interactions memory-intensive. During listening, the model can finish processing an audio unit before the next arrives; we term the remaining interval listening-time slack. We propose acoustic-to-text KV compression, which introduces a transcription side channel to convert incoming speech into compact textual memory within this interval. When the cache exceeds a target budget during inference, older acoustic states are evicted while transcripts and recent acoustic context remain. We train the side channel with LoRA using cross-entropy on transcription segments. To preserve listening and speaking behavior, we apply knowledge distillation to the original model's token-level output distributions at native prediction positions. On ten-minute LongSpeech sessions, our MiniCPM-o 4.5 implementation reduces peak streaming KV-cache size by 64.6% compared with the same model without eviction. The proposed method also improves transcription, temporal question answering, and summarization over the baseline. Full-Duplex-Bench evaluations further show comparable pause-handling, turn-taking, and interruption performance. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.AI). Categorie arXiv: Sound (cs.SD) ; Artificial Intelligence (cs.AI); Audio and Speech Processing (eess.AS). Pagina del lavoro: https://arxiv.org/abs/2609.31224 — 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 2026-10-08T01:12:14.556Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).