INESC-ID RCAAP Portal
Not a member yet
15883 research outputs found
Sort by
Pasture resilience: phenological patterns and critical thresholds in the face of climate change
No full textMENTOR: Fixing introductory programming assignments with formula-based fault localization and LLM-driven program repair
No full textExploring features for membership inference in ASR model auditing
No full textThis study explores the effectiveness of loss-based features in combination with Gaussian and adversarial perturbations for membership inference (MI) in automatic speech recognition (ASR) models. The authors propose a novel approach to MI, leveraging loss information from ASR models, which outperforms existing error-based features at sample-level MI, particularly when combined with proposed perturbations. By examining the impact of different feature sets and levels of access to target models, this work provides valuable insights for auditing ASR systems and highlights the importance of considering various factors in effective MI
Citizen participation in flood risk management: Usability Insights from the RiverCure Portal
No full text
