DEEPSE Forum Seminars - Bridging the Semantic Gap: From Design-time Software Design & Testing to Runtime Explainability

Speaker:  Prof. Jialong Li

13 Febbraio 2026 | 14:30
DEIB - Aula BIO1 (Ed. 21)
On line by Teams

Contatti: Bruno Guindani

Sommario

On February 13^th, 2026, at 2:30 pm will we have a new seminar of DEEPSE Forum series titled "Bridging the Semantic Gap: From Design-time Software Design & Testing to Runtime Explainability".

Modern software-intensive systems face a dual challenge: the "internal" difficulty of designing and testing complex logic and the "external" difficulty of making system behavior understandable to humans. We approach both challenges from the perspective of the "Semantic Gap"—the disconnect between low-level system execution and high-level human intent. In this talk, I will present our recent efforts on narrowing this gap from two perspectives:
For internal difficulty in software design and testing, we explore how LLMs can enhance SBSE by incorporating semantic awareness. This includes automating the generation and optimization of ML models, system representations, and test cases through semantic mapping.
For external explainability, we explored a series of explainability techniques (including interface design and the interpretation of software analysis results) to align human understanding with system logic or perception. These domains include autonomous driving, Human-LLM interaction, and Android privacy mechanisms.

Biografia

Jialong Li is a fixed-term Assistant Professor at the Waseda Institute for Advanced Study (WIAS), Waseda University, Japan. He received his Ph.D. in March 2025 under the supervision of Prof. Kenji Tei, with a research focus on Discrete Controller Synthesis and graceful degradation. During his doctoral studies, he was a visiting student in the groups of Prof. Zhi Jin and Prof. Eunsuk Kang. He has also worked with Prof. Christos Tsigkanos on F Prime space software. His current research interests include self-adaptive systems, SBSE, and Human-Software Interaction (specifically explainability and transparency).