Manacá-1B-Instruct-2: Reproducible Safety Alignment for a Native Brazilian-Portuguese LLM. Negative DPO Results, a Controlled Composition Scan, and a Silent Failure Mode of the LLM Judge
National Laboratory for Scientific Computing (LNCC)(国家科学计算实验室);Institute of Artificial Intelligence(人工智能研究所)
Bruno Leonardo Santos Menezes, Carlos Leonardo Souza Cardoso, Fabio André Machado Porto
Brazilian Portuguese remains underserved by open-source language models, and the few that are safety-aligned rarely document what did not work or validate comparisons with statistical significance. We present Manacá-1B-Instruct-2, the second-generation post-training alignment of Manacá-1B (1.72B parameters, trained from scratch for PT-BR), with an SFT corpus of 291,223 examples. We document three DPO attempts, all negative, including a case where the training reward margin reaches 1.0 while free generation does not change at all, and show that safety came from a dedicated safety-SFT (LLM-jp AnswerCarefully style), refined over three iterations until closing the safety-helpfulness frontier. A controlled scan of the fraction of general data in the safety-SFT mix (10% to 60%) reveals a non-monotonic over-refusal curve with a knee point at 40-50%. The final model is compared against four external PT-BR models on the same items, with Wilson confidence intervals and paired McNemar tests with Holm-Bonferroni correction: the lower bound of its safety interval (94.1%) lies above the upper bound of all four baselines. We also report a methodological finding for LLM-judge evaluation: on two of six safety items in MT-Bench-PT, the judge categorically refuses to assign a score regardless of the evaluated response, silently reducing the effective N of that category in every prior report of the project. We complement this with an n-gram decontamination check and a Chinchilla-anchored compute-budget table that attributes the gap in general capability to pre-training scale rather than alignment methodology. All code, raw data, logs, and scripts that produce every reported number are publicly released.
Zero-shot Performance of Generative AI in Brazilian Portuguese Medical Exam
Einstein Global Advanced Technologies for Equity(埃因斯坦全球先进科技以公平为宗旨);Hospital Israelita Albert Einstein(埃因斯坦医院);Departamento de Pacientes Graves(重症患者部门);Stanford Center for Artificial Intelligence in Medicine and Imaging(斯坦福大学医学与成像人工智能中心);Departmento de Cirurgia(外科部门);Faculdade de Medicina, Universidade de São Paulo(圣保罗大学医学院);Faculdade Israelita de Ciências da Saúde Albert Einstein(埃因斯坦以色列健康科学学院)
本文在 HCFMUSP 巴西葡语医学住院医师考试零样本条件下评测六个 LLM 与四个 MLLM,发现顶级 Claude 模型准确率接近人类,但多模态图像题仍有差距。
RDF2PT: Generating Brazilian Portuguese Texts from RDF Data
University of Leipzig(莱比锡大学);Tilburg University(蒂尔堡大学);University of Wolverhampton(伍尔弗汉普顿大学);Military Institute of Engineering - Brazil(巴西军事工程学院);Federal University of Rio de Janeiro - Brazil(巴西里约热内卢联邦大学);German Federal Institute for Risk Assessment - Germany(德国联邦风险评估研究所);Paderborn University(帕德博恩大学)
Brazilian Portuguese Speech Recognition Using Wav2vec 2.0
Federal University of Technology - Paraná(巴拉那联邦理工大学);University of São Paulo(圣保罗大学);Federal University of Mato Grosso(马托格罗索联邦大学);Federal University of Goias(戈亚斯联邦大学)
TTS-Portuguese Corpus: a corpus for speech synthesis in Brazilian Portuguese
Federal University of Technology – Paraná(巴拉那联邦理工大学);DefinedCrowd Corp.(DefinedCrowd公司);Federal University of Mato Grosso(马托格罗索联邦大学);Research Center in Digitalization and Intelligent Robotics (CEDRI)(数字化与智能机器人研究中心(CEDRI))
NILC-Metrix: assessing the complexity of written and spoken language in Brazilian Portuguese
Instituto de Ciências Matemáticas e de Computação - University of São Paulo(圣保罗大学数学与计算机科学研究所);The University of Sheffield(谢菲尔德大学);Itaú Unibanco(伊塔乌联合银行)
Recognizing the vocabulary of Brazilian popular newspapers with a free-access computational dictionary
Federal University of Rio Grande do Sul (UFRGS)(南里奥格兰德联邦大学(UFRGS));Federal University of São Carlos (UFSCar)(圣卡洛斯联邦大学(UFSCar));Université Paris-Est (UPE)(巴黎东大学(UPE))
A Large-Scale Dataset for Multi-Class Financial News Classification in Brazilian Portuguese
Federal University of Paraíba(帕拉伊巴联邦大学);Federal University of Campina Grande(帕坎纳联邦大学);Federal Rural University of Pernambuco(伯南布哥联邦农村大学);Federal University of Pernambuco(伯南布哥联邦大学)
Carolina: a General Corpus of Contemporary Brazilian Portuguese with Provenance, Typology and Versioning Information
University of São Paulo(圣保罗大学);Federal Institute of Education, Science and Technology of Bahia(巴伊亚联邦教育、科学与技术学院);State University of Southwestern Bahia(巴伊亚西南州立大学)
Enhancing Portuguese Variety Identification with Cross-Domain Approaches
Faculty of Sciences, University of Porto(波尔图大学理学院);Faculty of Engineering, University of Porto(波尔图大学工程学院);INESC TEC;Faculty of Arts and Humanities, University of Porto(波尔图大学艺术与人文学院);Centre of Linguistics, University of Porto(波尔图大学语言学中心);University of Beira Interior(贝拉内瓦尔大学);Ci2 - Smart Cities Research Center(Ci2智慧城市研究中心)
AI-PAVE-Br: Leveraging Large Language Models for Enhanced Product Attribute Value Extraction through a Golden Set Approach
LuizaLabs – Center of Excellence in Artificial Intelligence(LuizaLabs人工智能卓越中心);Department of Computing and Informatics – Mackenzie Presbyterian University(计算与信息学系-麦金西 Presbyterian大学);Institute of Mathematics and Computer Sciences - University of São Paulo(数学与计算机科学研究所-圣保罗大学)