Institution profile

Fondazione Bruno Kessler

Academic institutioneurope · it
Official website
Research library285linked papers
Opportunities0open roles
Selected work

Representative Papers

LeWiDi-2025 at NLPerspectives: The Third Edition of the Learning with Disagreements Shared Task

Oct 09, 2025

This work addresses the challenge of modeling and evaluating AI systems’ capacity to capture human judgment variability—such as disagreement and subjectivity. Methodologically, we (1) extend the LeWiDi benchmark to four tasks (paraphrase identification, irony/sarcasm detection, natural language inference) with ordinal annotations and individual-perspective prediction; (2) introduce the first integration of soft-label learning and annotator modeling, moving beyond hard-classification paradigms; and (3) propose a multi-task training framework jointly optimizing distributional prediction, individual annotator modeling, and population-level judgment distribution learning. Contributions include two novel evaluation metrics that surpass conventional measures like cross-entropy, and comprehensive empirical analysis revealing strengths and limitations of existing approaches in modeling judgment variability. These advances significantly enhance LeWiDi’s utility and extensibility as a benchmark platform for controversy-aware AI.

2 citationsRead paper

Exploiting Assumptions for Effective Monitoring of Real-Time Properties under Partial Observability

Sep 09, 2024IEEE International Conference on Software Engineering and Formal Methods

Runtime verification of temporal properties—such as those expressed in Metric Interval Temporal Logic (MITL)—is challenging in partially observable real-time systems, particularly when critical internal events (e.g., latent faults) remain unobservable. Method: This paper proposes an active prediction approach grounded in prior formal assumptions: system behavior is modeled as a timed automaton and integrated into a runtime verification framework to enable online inference of unobservable internal events. For the first time, formal system assumptions are deeply embedded into the real-time monitoring pipeline, combining constraint-driven temporal observation modeling with assumption-guided online verification. The approach is implemented within the UPPAAL toolchain. Contribution/Results: Experimental evaluation demonstrates that the method predicts property satisfaction/violation up to several time units in advance. In case studies involving smart grids and medical devices, monitoring success rates for properties dependent on unobservable events improve by 47%, significantly enhancing both the foresight and completeness of runtime verification.

2 citationsRead paper
Recent publications

Latest Papers

FineHOI: Part-Aware Dense Representations for Zero-Shot Human-Object Interaction Detection

Sep 05, 2026

Human-Object Interaction (HOI) detection aims to localize humans and objects in images and classify their interactions. Zero-shot HOI focuses on recognizing interactions that are not observed during training, requiring models to generalize beyond seen verb-object compositions. Recent approaches leverage Vision-Language Models (VLMs), benefiting from rich semantic representations. However, they often rely on global or detector-centric features that compress interaction cues and hinder fine-grained spatial reasoning. To overcome this limitation, we propose FineHOI, a zero-shot HOI framework that explicitly models interactions from dense patch-level features. Our approach is motivated by the observation that human-object interactions are defined by localized spatial relationships, which are not preserved by global and detector-centric representations. To this end, we introduce an Adaptive Part-Level Attention module that decomposes humans and objects into semantically coherent parts via unsupervised clustering, and re-weights them based on their interaction relevance. These representations are then integrated through a Region-Aware Interaction Transformer that integrates part-aware and global features and produces the final HOI embedding. Extensive experiments demonstrate that FineHOI consistently outperforms existing zero-shot HOI methods, achieving particularly strong gains on unseen interactions. Code is available at https://github.com/francescotonini/fine-hoi.

0 citationsRead paper