Our research group at the forefront of mathematical logic —particularly non-classical and substructural logics— is uniquely positioned to drive innovation through technology transfer. By offering cutting-edge educational services in logic instruction, custom-designed benchmarks for large language models (LLMs), and sophisticated tools for argument analysis and rational disagreement modeling, the group empowers organizations to achieve new levels of precision and interpretability in AI development. With deep theoretical expertise and a commitment to practical impact, the group delivers solutions that enhance critical reasoning capacities, optimize the evaluation of intelligent systems, and pave the way for the next generation of trustworthy and transparent AI technologies.
We invite industry partners, educational institutions, and AI developers to collaborate with us in transforming advanced logic research into powerful, real-world solutions.
Director: Ignacio Mastroleo (University of Buenos Aires and CONICET)
Academic coordinator: Eduardo Barrio (University of Buenos Aires and CONICET)
Research led: Federico (University of Tümbingen and CONICET)
Funded by: ARIA
Researching potential Earth cooling approaches raises profound ethical and societal questions that require careful consideration and robust governance frameworks, especially ensuring diverse global perspectives are included. This project focuses on building research capacity and developing ethical frameworks, particularly within the Global South. This project will build a Latin America/Caribbean-UK research network that will address fundamental questions regarding the governance of these approaches, as well as nurturing a new community of experts in the region. The work will explore societal implications, ethics frameworks for managng trade-offs and the breadth of opinions, co-production of knowledge and regional governance, particularly in the Latin America/Caribbean context.
Directors: Eduardo Barrio (University of Buenos Aires and CONICET)
Funded by: University of Buenos Aires
Despite the rapid and unexpected advancement of new Large Language Models (LLMs), there is a marked deficit in the development of standardized assessments, or benchmarks, for these LLMs. This deficit is even greater in benchmarking for general language comprehension or reasoning skills. Currently, the service sector is seeking to integrate these LLMs into the production system, automating some processes. However, this entails certain risks, given that current benchmarks have little representation of languages such as Spanish and its variants, so the quality of LLMs will not be adequately assessed for local applications. To prevent problems when integrating LLMs into the local economy, it is necessary to develop benchmarks for comprehension of Spanish and its Rioplatense variant, as well as for reasoning skills that facilitate the generalization of LLM performance.
A website that includes mathematical logic exercises with automatic correction and solution.
See the interview with its creator Ariel Roffé and Eduardo Barrio, published by CONICET by clicking this link
Technology developer: Ariel Roffe (University of Buenos Aires and CONICET
As part of our educational services, we are proud to promote TAUT, an innovative platform for the teaching and learning of logic. Developed with a focus on clarity, precision, and accessibility, TAUT offers interactive tools for studying a wide range of logical systems, from classical logic to non-classical and substructural frameworks.
Designed for students, educators, and researchers alike, TAUT provides a dynamic environment for practicing proof construction, understanding formal semantics, and exploring logical structures intuitively and engagingly.
Through TAUT, BA-Logic, and BA-AI Lab deliver state-of-the-art educational resources that complement traditional instruction and foster deeper, more rigorous logical reasoning skills.
We invite academic institutions, training programs, and independent learners to integrate TAUT into their curricula and join us in building the next generation of critical thinkers.
Explore TAUT at taut-logic.com
An initiative to develop custom-designed benchmarks for large language models (LLMs), and sophisticated tools for argument analysis and rational disagreement modeling
BA-AI Lab is a pioneering initiative at the intersection of logic and artificial intelligence. BA-AI Lab is dedicated to advancing the evaluation and development of large language models (LLMs) through the design of adversarial benchmarks, with a particular emphasis on assessing reasoning capabilities. Our first benchmark targets the critical but often underexplored dimension of logical reasoning, offering rigorous, adversarially constructed test sets that expose the limitations and strengths of current LLM architectures.
By leveraging deep expertise in mathematical logic, formal semantics, and substructural logics, BA-AI Lab provides industry partners, research institutions, and AI developers with sophisticated evaluation tools that go beyond surface-level performance metrics. Our services are designed to promote the development of more reliable, interpretable, and robust AI systems, helping to ensure that reasoning processes within LLMs meet the highest standards of validity and rationality.
We invite collaborators to work with us in setting new benchmarks for the future of AI reasoning.