Metraipolis › archive › dc_c479f8cd3116
mu-bench: A Multilingual Utterance Transcription Benchmark — arXiv:2609.32082
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.32082
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T21:49:57.695Z
- SHA-256 of the archived text
8a873e96f3b2f40842a81e20f43465c354b101bcaa2efab4404e93b09a1110c3- Stable URL of this page
https://metraipolis.com/documento/dc_c479f8cd3116
How to cite
mu-bench: A Multilingual Utterance Transcription Benchmark — arXiv:2609.32082 — arXiv (en), https://arxiv.org/abs/2609.32082 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c479f8cd3116 (Metraipolis archive dc_c479f8cd3116, sha256 8a873e96f3b2f40842a81e20f43465c354b101bcaa2efab4404e93b09a1110c3)
BibTeX
@misc{dcc479f8cd3116,
title = {mu-bench: A Multilingual Utterance Transcription Benchmark — arXiv:2609.32082},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.32082},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c479f8cd3116; sha256 8a873e96f3b2f40842a81e20f43465c354b101bcaa2efab4404e93b09a1110c3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32082. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Voice agents depend on accurate automatic speech recognition (ASR) to act on what callers say, yet ASR is evaluated on read, English-centric speech with word error rate (WER), which penalizes surface rather than semantic differences. We introduce mu-bench, a dataset of 4,270 caller utterances from 250 phone calls to an AI banking agent in English, Spanish, Turkish, Vietnamese, and Mandarin, centered on form-field inputs such as names, email addresses, and confirmation codes. We release Utterance Error Rate (UER), an LLM judge of whether a transcript preserves meaning, calibrated against human raters, together with an LLM normalizer that makes WER comparable across providers' output formats. On 1,847 human-rated transcripts, UER agrees with annotators at $\kappa$ = 0.78, versus 0.53 for exact-match WER on normalized text. We rank six commercial providers on a public leaderboard; the best reaches 11.9% UER, and Mandarin is hardest for all six. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CL). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2609.32082 — 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 2026-10-07T21:27:08.389Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).