Datasets:

lusollm
/

bagaco4-split-edu-dev

license: cc-by-4.0
task_categories:
  - text-generation
language:
  - pt
tags:
  - pretraining
  - pt-pt
  - Portugal
  - bagaco-mix
  - arxiv:2603.03543
  - arxiv:2602.05879
  - arxiv:2503.10267
  - arxiv:2107.06499
  - arxiv:2201.06642
  - arxiv:1911.02116
pretty_name: Bagaco4, o maior corpo publico de pré-treino em português (pt-pt)
size_categories:
  - 10M<n<100M

Corpo público de pré-treino dedicado exclusivamente ao português de europeu (pt-PT).

O corpos foi curado de modo a procurar um bom alinhamento com o português usado em Portugal, e minimizar as contaminações de outras variantes da língua ou traduções automáticas de baixa qualidade.

A principal inovação aplicada ás várias versões do bagaco, consiste na classificação prévia do texto de todos os documentos com um score pt-pt, e extrair apenas os documentos que atingem o limiar que garante um score pt-pt superior a 0.90.

Esta versão está segmentada por níveis de qualidade. Este formato apresenta uma ligeira vantagem para a constituição de datasets de pré-treino (com ramping) adaptados ao modelo a treinar.

edu	pt-pt	docs	chars	words	tokens1	tokens2
5	0.8933	128,498	598,609,752	96,096,649	151 748 724	167,413,547
4	0.8958	2,574,687	19,598,337,214	3,165,154,084	4,788,631,045	4,303,521,535
3	0.9040	5,811,431	39,792,992,065	6,343,431,560	9,510,608,456	8,215,187,365
2	0.9290	13,738,836	53,091,805,170	8,557,967,628	12,391,659,157	12 156 495 527
1	0.9390	31,443,575	73,734,288,823	11,826,372,987	17,637,450,554	17,987,353,947
0	0.9184	44,850,283	120,304,072,247	19,438,743,284	30,155,035,328	32,662,463,995
Total	-	98,547,310	307,120,105,271	49,427,766,192	74 635 133 264	75,492,435,916

tokens1: nanochat rustbpe 32k dic tokens2: eurollm

Referências

Este corpo foi produzido pela seleção de vários corpos pelo que a utilização implica a boa prática de citar também os trabalhos originais:

Gigaverbo-v2

@misc{correa2026tucano2cool,
      title={{Tucano 2 Cool: Better Open Source LLMs for Portuguese}}, 
      author={Nicholas Kluge Corr{\^e}a and Aniket Sen and Shiza Fatimah and Sophia Falk and Lennard Landgraf and Julia Kastner and Lucie Flek},
      year={2026},
      eprint={2603.03543},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2603.03543}, 
}

EuroWeb-2512

@misc{ramos2026eurollm22btechnicalreport,
      title={EuroLLM-22B: Technical Report},
      author={Miguel Moura Ramos and Duarte M. Alves and Hippolyte Gisserot-Boukhlef and João Alves and Pedro Henrique Martins and Patrick Fernandes and José Pombal and Nuno M. Guerreiro and Ricardo Rei and Nicolas Boizard and Amin Farajian and Mateusz Klimaszewski and José G. C. de Souza and Barry Haddow and François Yvon and Pierre Colombo and Alexandra Birch and André F. T. Martins},
      year={2026},
      eprint={2602.05879},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2602.05879},
}

HPLT2.0_cleaned

@inproceedings{burchell-etal-2025-expanded,
    title = "An Expanded Massive Multilingual Dataset for High-Performance Language Technologies ({HPLT})",
    author = {Burchell, Laurie  and
      de Gibert, Ona  and
      Arefyev, Nikolay  and
      Aulamo, Mikko  and
      Ba{\~n}{\'o}n, Marta  and
      Chen, Pinzhen  and
      Fedorova, Mariia  and
      Guillou, Liane  and
      Haddow, Barry  and
      Haji\({\v{c}},\) Jan  and
      Helcl, Jind\({\v{r}}\)ich  and
      Henriksson, Erik  and
      Klimaszewski, Mateusz  and
      Komulainen, Ville  and
      Kutuzov, Andrey  and
      Kyt{\"o}niemi, Joona  and
      Laippala, Veronika  and
      M\({\ae}\)hlum, Petter  and
      Malik, Bhavitvya  and
      Mehryary, Farrokh  and
      Mikhailov, Vladislav  and
      Moghe, Nikita  and
      Myntti, Amanda  and
      O{'}Brien, Dayy{\'a}n  and
      Oepen, Stephan  and
      Pal, Proyag  and
      Piha, Jousia  and
      Pyysalo, Sampo  and
      Ram{\'i}rez-S{\'a}nchez, Gema  and
      Samuel, David  and
      Stepachev, Pavel  and
      Tiedemann, J{\"o}rg  and
      Vari\({\v{s}},\) Du\({\v{s}}\)an  and
      Vojt\({\v{e}}\)chov{\'a}, Tereza  and
      Zaragoza-Bernabeu, Jaume},
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.acl-long.854/",
    doi = "10.18653/v1/2025.acl-long.854",
    pages = "17452--17485",
    ISBN = "979-8-89176-251-0",
}

CrawlPT_dedup

@inproceedings{garcia-etal-2024-robertalexpt,
    title = "{R}o{BERT}a{L}ex{PT}: A Legal {R}o{BERT}a Model pretrained with deduplication for {P}ortuguese",
    author = "Garcia, Eduardo A. S.  and
      Silva, Nadia F. F.  and
      Siqueira, Felipe  and
      Albuquerque, Hidelberg O.  and
      Gomes, Juliana R. S.  and
      Souza, Ellen  and
      Lima, Eliomar A.",
    editor = "Gamallo, Pablo  and
      Claro, Daniela  and
      Teixeira, Ant{\'o}nio  and
      Real, Livy  and
      Garcia, Marcos  and
      Oliveira, Hugo Gon\({\c{c}}\)alo  and
      Amaro, Raquel",
    booktitle = "Proceedings of the 16th International Conference on Computational Processing of Portuguese",
    month = mar,
    year = "2024",
    address = "Santiago de Compostela, Galicia/Spain",
    publisher = "Association for Computational Lingustics",
    url = "https://aclanthology.org/2024.propor-1.38",
    pages = "374--383",
}