Dataset · Hugging Face

Greenbean/RoleBreak

Name: RoleBreak
Creator: Greenbean
Keywords: task_categories:audio-text-to-text, task_categories:text-to-speech, language:en, size_categories:n<1K, format:json, modality:text, library:datasets, library:pandas, library:polars, library:mlcroissant, arxiv:2609.16614, region:us, role-playing, speech-to-speech, benchmark

RoleBreak A benchmark for long-horizon role-playing robustness in spoken dialogue. RoleBreak holds 310 roles, 6,688 human-verified user turns (21.6 per conversation) and 11,743 fine-grained pass/fail criteria. Each conversation puts a speech-to-speech model in character and then stresses it as context accumulates — context-dependent probes and targeted interventions against role consistency, interaction quality, safety, and affect. This repository includes three things: The… See the full description on the dataset page: https://huggingface.co/datasets/Greenbean/RoleBreak.

View on Hugging FaceReport an issue

Author / organization	Greenbean
Downloads	1.1K (as reported by Hugging Face)
Likes	1
Created	23 August 2026
Last updated at source	16 September 2026

role-playingspeech-to-speechbenchmarklong-horizonevaluation

Embed this entry

<iframe src="https://silytics.si/embed/item/22e8a6b4-2bd8-451f-84b3-340c81d0c962" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

Popularity, last 2 days

1,105 now (-5 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.

Related papers

RoleBreak: Benchmarking Long-Horizon Role-Playing Robustness in Spoken DialoguePaper

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.