Image: Fundação para a Ciência e a Tecnologia (FCT)
INESC-ID secures two AI grants from CMU Portugal and UT Austin Portugal: improving LLM code generation and humanoid domestic robots
INESC-ID has been awarded two research grants under the 2025 exploratory project calls of the CMU Portugal and UT Austin Portugal programmes, funded by Fundação para a Ciência e a Tecnologia (FCT). one focused on making LLM-generated code more reliable, the other on enabling humanoid robots to reason and communicate naturally with humans.
Selected from 36 applications to the CMU Portugal call and 41 to the UT Austin Portugal call, the two projects were among a total of 15 approved across both programmes.
José Fragoso Santos leads the project Quantitative Program Analysis for Improving LLM Code Generation, which addresses a growing risk in software development: LLMs are increasingly deployed for coding tasks in industrial settings, yet they still generate code with subtle bugs, basic reasoning failures, and security vulnerabilities. The project will investigate a new approach, Reinforcement Learning from Static Analysis Feedback (RLSAF), which trains LLMs to account for the semantic properties of the programs they produce, rather than treating programming as a natural language task. This approach has the potential to help programmers produce better and more trustworthy software.
Bruno Martins leads the project EXPlanation and ObseRvation for HumanOid DOmestic RObots (EXPLOR-HODOR), which aims to develop an LLM-based AI system for robot cognition and interaction, specialised in domestic tasks and collaboration with humans. The project focuses on two challenges: investigating how LLM-based models can help robots reason about their environment and explain their decisions, and enabling more natural spoken interaction with humans, moving beyond turn-based dialogue. Both will be tested in realistic home settings, including participation in the RoboCup@Home competition.
The two projects were selected under FCT’s 2025 exploratory calls that prioritise Information Technologies (CMU Portugal) and Advanced Computing (UT Austin Portugal). Project 2025.16411.CMU received €70,000; project 2025.17632.UTA received €50,000. Duration: 12 months.
—
© 2026 INESC-ID. When sharing or adapting this article, credit INESC-ID with a link to the original source.
Image: © 2026 Fundação para a Ciência e a Tecnologia (FCT)