How machines learn,
studied in Siena.
Theories of learning in structured domains, put to work on large language models, vision, biology and crosswords.
See the nine research linesDIISM, University of Siena, Via Roma 56, 53100 Siena, Italy
The Palazzo Pubblico and the Torre del Mangia, drawn as a network. A signal climbs from the door to the bell once, when the page opens.
Nine ways of asking how machines learn
The Campo of Siena is divided into nine sections, after the Nine who governed the city. Our research has nine lines. Choose one.
Piazza del Campo seen from above: the Palazzo Pubblico is the black bar at the bottom.
Learning from constraints
Animals coordinate perception with action, and children move from sensorimotor to abstract representations in stages. This line asks whether such behaviour follows general principles of learning, and models knowledge as constraints that a learning agent must satisfy. The book Machine Learning: A Constraint-Based Approach (Marco Gori) grew out of it.
News
- Multi-SFT: Permissively-Licensed Multilingual Instruction-Following
- TalkFa: A Unified Benchmark for Farsi Dialogue Generation and Understanding
- From Graph to Text and Back: Semantic Fidelity in Automated Industrial Knowledge Graphs
- MultiHint-9K: Orchestrating Human and AI Agency for Scalable Pedagogical Hint Generation
Seminars
- >>> Website on Collectionless AI – Open! – https://collectionless.ai/ <<<
- [Nov 15th 2023] Applications of Large Language Models in the field of education
- External Seminar On Bayesian Inference and Generative Models for Wireless Communication Algorithms Organised by prof. Andrea Abrardo [Apr 12th, 2023 – 10:00 – 13:00 AM]
- [Apr 5th 2023] The expressive power of pooling in Graph Neural Networks