Zobrazeno 1 - 8
of 8
pro vyhledávání: '"Tomani, Christian"'
A major barrier towards the practical deployment of large language models (LLMs) is their lack of reliability. Three situations where this is particularly apparent are correctness, hallucinations when given unanswerable questions, and safety. In all
Externí odkaz:
http://arxiv.org/abs/2404.10960
Autor:
Tomani, Christian, Vilar, David, Freitag, Markus, Cherry, Colin, Naskar, Subhajit, Finkelstein, Mara, Garcia, Xavier, Cremers, Daniel
Maximum-a-posteriori (MAP) decoding is the most widely used decoding strategy for neural machine translation (NMT) models. The underlying assumption is that model probability correlates well with human judgment, with better translations getting assig
Externí odkaz:
http://arxiv.org/abs/2310.06707
Calibrating deep learning models to yield uncertainty-aware predictions is crucial as deep neural networks get increasingly deployed in safety-critical applications. While existing post-hoc calibration methods achieve impressive results on in-domain
Externí odkaz:
http://arxiv.org/abs/2302.05118
Given the importance of getting calibrated predictions and reliable uncertainty estimations, various post-hoc calibration methods have been developed for neural networks on standard multi-class classification tasks. However, these methods are not wel
Externí odkaz:
http://arxiv.org/abs/2210.06391
Autor:
Tomani, Christian, Cremers, Daniel
We show that utilizing attribution maps for training neural networks can improve regularization of models and thus increase performance. Regularization is key in deep learning, especially when training complex models on relatively small datasets. In
Externí odkaz:
http://arxiv.org/abs/2205.15094
We address the problem of uncertainty calibration and introduce a novel calibration method, Parametrized Temperature Scaling (PTS). Standard deep neural networks typically yield uncalibrated predictions, which can be transformed into calibrated confi
Externí odkaz:
http://arxiv.org/abs/2102.12182
Autor:
Tomani, Christian, Gruber, Sebastian, Erdem, Muhammed Ebrar, Cremers, Daniel, Buettner, Florian
We address the problem of uncertainty calibration. While standard deep neural networks typically yield uncalibrated predictions, calibrated confidence scores that are representative of the true likelihood of a prediction can be achieved using post-ho
Externí odkaz:
http://arxiv.org/abs/2012.10988
Autor:
Tomani, Christian, Buettner, Florian
To facilitate a wide-spread acceptance of AI systems guiding decision making in real-world applications, trustworthiness of deployed models is key. That is, it is crucial for predictive models to be uncertainty-aware and yield well-calibrated (and th
Externí odkaz:
http://arxiv.org/abs/2012.10923