Publications

Welcome to the Deliverables & Documents section! Here you will find project deliverables published on Zenodo as well as other written outputs by the project.
Stay tuned for updates!

Type of content
Showing 11 - 20 of 41
Conference paper

Towards Transparent Reasoning: What Drives Faithfulness in Large Language Models?

Published on:
Large Language Models (LLMs) often produce explanations that do not faithfully reflect the factors driving their predictions. In healthcare settings, such unfaithfulness is especially problematic: explanations that omit salient clinical cues or mask spurious shortcuts can undermine clinician trust and lead to unsafe decision support. ... Read more
Authors: Università della Svizzera italiana
Conference paper

Deferring Concept Bottleneck Models: Learning to Defer Interventions to Inaccurate Experts

Published on:
Concept Bottleneck Models (CBMs) are interpretable machine learning models that ground their predictions on human-understandable concepts, allowing for targeted interventions in their decision-making process.... Read more
Authors: Università della Svizzera italiana
Scientific publication

Navigating multi-country, multi-disciplinary co-design with families, adolescents and healthcare practitioners: the SmartCHANGE project.

Published on:
Early prevention of non-communicable diseases (NCDs) by improving lifestyle behaviors such as diet and physical activity among children and adolescents has significant potential to reduce the burden of disease and increase their quality of life.... Read more
Authors: Amsterdam UMC Location VUmc
Project deliverable

D6.2 - Final technical specification

Published on:
This document presents the final and updated technical specification for Work Package 6 (WP6) as part of the SmartCHANGE project. It builds upon the previous version, D6.1: Initial Technical Specification, delivered at M18 (October 2024).... Read more
Authors: Engineering Ingegneria Informatica SpA
Project deliverable

D8.3 - Midterm dissemination and communication report and plan

Published on:
This midterm deliverable provides an overview of communication, dissemination, and stakeholder engagement activities carried out during the first half of the SmartCHANGE project, alongside a forward-looking strategy for the remainder of the project.... Read more
Authors: Trust-IT Services
Conference paper

Avoiding Leakage Poisoning: Concept Interventions Under Distribution Shifts

Published on:
In this paper, we investigate how concept-based models (CMs) respond to out-of-distribution (OOD) inputs. CMs are interpretable neural architectures that first predict a set of high-level concepts (e.g., "stripes", "black") and then predict a task label from those concepts.... Read more
Authors: Università della Svizzera italiana
Conference paper

Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning

Published on:
Causal opacity denotes the difficulty in understanding the "hidden" causal structure underlying the decisions of deep neural network (DNN) models. This leads to the inability to rely on and verify state-of-the-art DNN-based systems, especially in high-stakes scenarios.... Read more
Authors: Università della Svizzera italiana
Conference paper

Counterfactual Concept Bottleneck Models

Published on:
Current deep learning models are not designed to simultaneously address three fundamental questions: predict class labels to solve a given classification task (the "What?"), simulate changes in the situation to evaluate how this impacts class predictions (the "How?"), and imagine how the scenario... Read more
Authors: Università della Svizzera italiana
Conference paper

AnyCBMs How to Turn Any Black Box into a Concept Bottleneck Model

Published on:
Originally published by CEUR for the XAI World Conference 2024.... Read more
Authors: Università della Svizzera italiana
Conference paper

Health Risk Prediction for Children Built on a Combination of Heterogeneous Datasets

Published on:
... Read more
Authors: Jožef Stefan Institute