Seminari
Da febbraio 2019, l’Istituto del Software ha iniziato una serie di seminari. Ogni giovedì pomeriggio, un ricercatore dell’Istituto terrà un breve discorso su un argomento di ingegneria del software a sua scelta, come ad esempio articoli interessanti pubblicati di recente, articoli seminali nel proprio campo di ricerca, discussioni su idee preliminari, tutorial e piccoli esperimenti.
Sulla nostra playlist YouTube potete riguardare alcuni dei seminari precedenti. Di seguito trovate ulteriori dettagli sul prossimo seminario, su quelli a venire e un archivio dei relatori precedenti.
Tutti sono invitati a partecipare ai seminari organizzati dall’Istituto del Software.
Prossimo Relatore: Jinhan Kim
Retrieval-Augmented Generation (RAG) systems have become the default way we deploy LLMs over proprietary knowledge, yet the way we test them has not kept pace. Most existing frameworks score answers one query at a time against a fixed snapshot of the corpus, which is useful for benchmarking, but a long way from testing the system as a whole. In this talk I want to take a software-testing view of RAG systems and ask two questions:
First, is our test suite actually exercising the retriever, or are we just hitting the same popular chunks over and over? I’ll introduce Chunk Coverage, an oracle-independent adequacy criterion that treats the corpus the way code coverage treats the program and show how it can guide test generation into regions of the retrieval space no existing tests have looked at.
Second, what happens when the corpus moves under your feet: updates, stale versions, OCR noise, format drift? A fixed <query, answer> pair is not a reliable oracle in that world. So I’ll talk about metamorphic testing for RAG: mutation operators that perturb the system at both the index and retrieved-context levels, and metamorphic relations that serve as a relational oracle in place of fixed ground truth.
Jinhan Kim is a postdoc at TAU lab in Università della Svizzera italiana (USI) led by Prof. Paolo Tonella. He completed his Ph.D. degree from KAIST under the supervision of Prof. Shin Yoo. His research bridges Software Engineering (SE) and Artificial Intelligence (AI), focusing on testing and reliability of AI systems. He develops principled methods to assess and improve the robustness of complex systems, from traditional software to LLMs, aiming to make them reliable and transparent for deployment in safety-critical domains. More information is available at https://jinhan.me/
Programma
-
7 Maggio 2026
-
21 Maggio 2026
-
28 Maggio 2026