terminal of open science

A novel sensor-fusion prognostics framework distinguishes between multiple failure modes in complex systems without manual inspection by clustering sensor data, enabling accurate remaining useful life prediction per mode. [arXiv:2411.12159]

Drift-Resilient TabPFN models temporal distribution shifts by learning Bayesian inference on causal models that gradually shift over time, improving accuracy by 8% and ROC AUC by 5.9% compared to state-of-the-art baselines on tabular data. [arXiv:2411.10634]