Thales Sales Almeida
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Thales Sales Almeida has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Giovana Kerche Bonás
Instituto de Computac ¸ão (IC) Universidade Estadual de Campinas (UNICAMP) Campinas -SP Brazil
9 shared publications - Rodrigo Nogueira
State University of Campinas (UNICAMP)
9 shared publications - Thiago Laitz
UNICAMP
8 shared publications - Hugo Abonizio
Maritaca AI
5 shared publications - Ramon Pires
Maritaca AI
5 shared publications - Celio Larcher
Jusbrasil
4 shared publications
Latest Publications
- 3 October 2026
- Published by Springer Nature
- p. 108-122
Prosa: Rubric-Based Evaluation of LLMs on Real User Chats in Brazilian Portuguese
- 3 October 2026
- Published by Springer Nature
- p. 108-122
Prosa: Rubric-Based Evaluation of LLMs on Real User Chats in Brazilian Portuguese
- 2 October 2026
- Published by Springer Nature
- p. 315-328
Do LLMs Take Sides? Measuring Opinion Bias and Sycophancy via Multi-turn Probing
- 2 October 2026
- Published by Springer Nature
- p. 315-328
Do LLMs Take Sides? Measuring Opinion Bias and Sycophancy via Multi-turn Probing
- 2 October 2026
- Published by Springer Nature
- p. 393-407
BLUEX V2: Benchmarking LLMs on Open-Ended Questions from Brazilian University Entrance Exams
- 2 October 2026
- Published by Springer Nature
- p. 393-407
BLUEX V2: Benchmarking LLMs on Open-Ended Questions from Brazilian University Entrance Exams
- 30 September 2026
- Published by Springer Nature
- p. 460-472
MARCA: A Checklist-Based Benchmark for Multilingual Web Search
- 30 September 2026
- Published by Springer Nature
- p. 460-472
MARCA: A Checklist-Based Benchmark for Multilingual Web Search
- 30 September 2026
- Published by Springer Nature
- p. 274-288
moBERTo: A Modern Encoder for Portuguese via Continued Pretraining of ModernBERT
- 30 September 2026
- Published by Springer Nature
- p. 274-288
moBERTo: A Modern Encoder for Portuguese via Continued Pretraining of ModernBERT
- 30 September 2026
- Published by Springer Nature
- p. 368-381
CAPITU: A Benchmark for Evaluating Instruction-Following in Brazilian Portuguese with Literary Context
- 30 September 2026
- Published by Springer Nature
- p. 368-381
CAPITU: A Benchmark for Evaluating Instruction-Following in Brazilian Portuguese with Literary Context
- 29 September 2025
- Published by Sociedade Brasileira de Computação
- p. 867-878
BLUEX Revisited: Enhancing Benchmark Coverage with Automatic Captioning
With the growing capabilities of Large Language Models (LLMs), there is an increasing need for robust evaluation methods, especially in multilingual and non-English contexts. We present an updated version of the BLUEX dataset, now including 2024-2025 exams and automatically generated image captions using state-of-the-art models, enhancing its relevance for data contamination studies in LLM pretraining. Captioning strategies increase accessibility to text-only models by more than 40%, producing 1,422 usable questions, more than doubling the number in the original BLUEX. We evaluated commercial and open-source LLMs and their ability to leverage visual context through captions.
- 29 September 2025
- Published by Sociedade Brasileira de Computação
- p. 867-878
BLUEX Revisited: Enhancing Benchmark Coverage with Automatic Captioning
With the growing capabilities of Large Language Models (LLMs), there is an increasing need for robust evaluation methods, especially in multilingual and non-English contexts. We present an updated version of the BLUEX dataset, now including 2024-2025 exams and automatically generated image captions using state-of-the-art models, enhancing its relevance for data contamination studies in LLM pretraining. Captioning strategies increase accessibility to text-only models by more than 40%, producing 1,422 usable questions, more than doubling the number in the original BLUEX. We evaluated commercial and open-source LLMs and their ability to leverage visual context through captions.
- 30 January 2025
- Published by Springer Nature
- p. 431-444
SurveySum: A Dataset for Summarizing Multiple Scientific Articles into a Survey Section
- 30 January 2025
- Published by Springer Nature
- p. 431-444
SurveySum: A Dataset for Summarizing Multiple Scientific Articles into a Survey Section
- Open Access
- 17 January 2025
- Published by Sociedade Brasileira de Computação in Journal of the Brazilian Computer Society
- Vol. 31 (1) , 1247-1263
Building High-Quality Datasets for Portuguese LLMs: From Common Crawl Snapshots to Industrial-Grade Corpora
The performance of large language models (LLMs) is deeply influenced by the quality and composition of their training data. While much of the existing work has centered on English, there remains a gap in understanding how to construct effective training corpora for other languages. We explore scalable methods for building web-based corpora for LLMs. We apply them to build a new 120B token corpus in Portuguese that achieves competitive results to an industrial-grade corpus. Using a continual pretraining setup, we study how different data selection and preprocessing strategies affect LLM performance when transitioning a model originally trained in English to another language. Our findings demonstrate the value of language-specific filtering pipelines, including classifiers for education, science, technology, engineering, and mathematics (STEM), as well as toxic content. We show that adapting a model to the target language leads to performance improvements, reinforcing the importance of high-quality, language-specific data. While our case study focuses on Portuguese, our methods are applicable to other languages, offering insights for multilingual LLM development.
- 17 January 2025
- Published by Sociedade Brasileira de Computação in Journal of the Brazilian Computer Society
- Vol. 31 (1) , 1247-1263
Open AccessBuilding High-Quality Datasets for Portuguese LLMs: From Common Crawl Snapshots to Industrial-Grade Corpora
The performance of large language models (LLMs) is deeply influenced by the quality and composition of their training data. While much of the existing work has centered on English, there remains a gap in understanding how to construct effective training corpora for other languages. We explore scalable methods for building web-based corpora for LLMs. We apply them to build a new 120B token corpus in Portuguese that achieves competitive results to an industrial-grade corpus. Using a continual pretraining setup, we study how different data selection and preprocessing strategies affect LLM performance when transitioning a model originally trained in English to another language. Our findings demonstrate the value of language-specific filtering pipelines, including classifiers for education, science, technology, engineering, and mathematics (STEM), as well as toxic content. We show that adapting a model to the target language leads to performance improvements, reinforcing the importance of high-quality, language-specific data. While our case study focuses on Portuguese, our methods are applicable to other languages, offering insights for multilingual LLM development.
- Open Access
- 17 January 2025
- Published by Sociedade Brasileira de Computação in Journal of the Brazilian Computer Society
- Vol. 31 (1) , 1078-1088
BRoverbs - Measuring how much LLMs understand Portuguese proverbs
Large Language Models (LLMs) exhibit significant performance variations depending on the linguistic and cultural context in which they are applied. This disparity signals the necessity of mature evaluation frameworks that can assess their capabilities in specific regional settings. In the case of Portuguese, existing evaluations remain limited, often relying on translated datasets that may not fully capture linguistic nuances or cultural references. Meanwhile, native Portuguese-language datasets predominantly focus on structured national exams or sentiment analysis of social media interactions, leaving gaps in evaluating broader linguistic understanding. To address this limitation, we introduce BRoverbs, a dataset specifically designed to assess LLM performance through Brazilian proverbs. Proverbs serve as a rich linguistic resource, encapsulating cultural wisdom, figurative expressions, and complex syntactic structures that challenge the model comprehension of regional expressions. BRoverbs aims to provide a new evaluation tool for Portuguese-language LLMs, contributing to advancing regionally informed benchmarking. The benchmark is available at https://huggingface.co/datasets/Tropic-AI/BRoverbs.
- 17 January 2025
- Published by Sociedade Brasileira de Computação in Journal of the Brazilian Computer Society
- Vol. 31 (1) , 1078-1088
Open AccessBRoverbs - Measuring how much LLMs understand Portuguese proverbs
Large Language Models (LLMs) exhibit significant performance variations depending on the linguistic and cultural context in which they are applied. This disparity signals the necessity of mature evaluation frameworks that can assess their capabilities in specific regional settings. In the case of Portuguese, existing evaluations remain limited, often relying on translated datasets that may not fully capture linguistic nuances or cultural references. Meanwhile, native Portuguese-language datasets predominantly focus on structured national exams or sentiment analysis of social media interactions, leaving gaps in evaluating broader linguistic understanding. To address this limitation, we introduce BRoverbs, a dataset specifically designed to assess LLM performance through Brazilian proverbs. Proverbs serve as a rich linguistic resource, encapsulating cultural wisdom, figurative expressions, and complex syntactic structures that challenge the model comprehension of regional expressions. BRoverbs aims to provide a new evaluation tool for Portuguese-language LLMs, contributing to advancing regionally informed benchmarking. The benchmark is available at https://huggingface.co/datasets/Tropic-AI/BRoverbs.