Current Projects

We develop multi-agentic systems to support Evidence-Based Medicine.

We develop methods to evaluate robustness, harmful prompt detection, and alignment dynamics in large language models.

We analyze coordination, cooperation, and emergent behavior in interacting agents at scales beyond individual human oversight.

We study fundamental questions in machine learning, such as learning dynamics, limits on representational capacity, and domain adaptation methods.

02

Medium

June 2026. Felippe Alves. AI agent harnesses. Medium
June 2026. Luiz Felipe Guain Teixeira. Learning at the Learning Conference: A Brief from ICLR 2026. Medium
May 2026. Thais Martins de Sousa. Emotion Concepts and their Function in a Large Language Model. Medium
April 2026. Luiz Felipe Guain Teixeira. Large Language Models for Systematic Reviews: What Multi-Agent Approaches Can Teach Us. Medium
March 2026. Eduardo Amorim. Robustness of Language Models against Harmful Prompts in an Underrepresented Language. Medium
March 2026. Felippe Alves. Emergent Misalignment: sneaky general misbehavior from a few specific bad examples. Medium

03

2026. L. F. G. Teixeira, C. P. de Barros, R. Vicente. Automated Risk of Bias Assessment in Systematic Reviews Using Large Language Models: A Multi-Model Single-Agent Benchmark on 227 Randomised Controlled Trials. preprint.
2026. F. Alves, R. Vicente. Kolmogorov-Arnold Networks for Small Language Models. preprint.
2026. E. A. Amorim, C. Zanchettin. Robustness of Language Models against Portuguese Harmful Prompts. IEEE WCCI 2026.
2026. D. B. Costa, R. Vicente. Persona-Model Collapse in Emergent Misalignment. arXiv:2605.12850.
2025. D. B. Costa, F. Alves, R. Vicente. Moral Susceptibility and Robustness under Persona Role-Play in Large Language Models. arXiv:2511.08565.
2025. D. B. Costa, R. Vicente. Deceive, Detect, and Disclose: Large Language Models Play Mini-Mafia. arXiv:2509.23023. AAMAS26.