Metraipolis › archive › dc_99e387745e2b
J-tau: A Fully Localized Japanese Benchmark for Tool-Agent-User Interaction — arXiv:2610.19638
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.19638
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-15T20:24:05.188Z
- SHA-256 of the archived text
e00a8fe946c4fb624ffaef9bf48443e6a4fe392f2c141b9a7c953088f657d459- Stable URL of this page
https://metraipolis.com/documento/dc_99e387745e2b
How to cite
J-tau: A Fully Localized Japanese Benchmark for Tool-Agent-User Interaction — arXiv:2610.19638 — arXiv (en), https://arxiv.org/abs/2610.19638 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_99e387745e2b (Metraipolis archive dc_99e387745e2b, sha256 e00a8fe946c4fb624ffaef9bf48443e6a4fe392f2c141b9a7c953088f657d459)
BibTeX
@misc{dc99e387745e2b,
title = {J-tau: A Fully Localized Japanese Benchmark for Tool-Agent-User Interaction — arXiv:2610.19638},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.19638},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_99e387745e2b; sha256 e00a8fe946c4fb624ffaef9bf48443e6a4fe392f2c141b9a7c953088f657d459},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.19638. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
We present J-tau, the first publicly available Japanese benchmark for evaluating tool-using LLM agents in interactive customer-support scenarios. While tau-bench is widely used to evaluate LLM agents, it is English-centric and cannot evaluate agents in other languages. Existing multilingual extensions offer no public Japanese version and rely primarily on machine translation, leaving locale-specific entities such as addresses and currencies inconsistent with the target region. To enable reliable evaluation in Japanese, we carefully localize tau-bench through machine translation, entity adaptation, and human review. During this process, we also uncover and fix two flaws in the original tau-bench: invalid tasks and a score aggregation error that inflates certain models' results. Experiments show that J-tau yields fewer evaluation errors than a naively machine-translated counterpart and that our fixes resolve both flaws. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2610.19638 — 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 2026-10-15T19:36:18.207Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).