Supervised fine-tuning corpus built from the verbatim records of the French
National Assembly. — Corpus de fine-tuning supervisé construit à partir des
comptes rendus intégraux de l'Assemblée nationale.
Each example is a contiguous block of a single sitting, pre-tokenised for
Qwen3.8-27B. The corpus covers
legislatures 15 to 17 (2017–2025) and is meant to teach a model to generate
parliamentary speech conditioned on speaker… See the full description on the dataset page:
https://huggingface.co/datasets/Codcordance/qwenmicycle-assnat-sft.
