When a deep learning model is deployed in the wild, it can encounter test data drawn from distributions different from the training data distribution and suffer drop in performance. For safe deployment, it is essential to estimate the accuracy of the pre-trained model on the test data. However, the labels for the test inputs are usually not immediately available in practice, and obtaining them can be expensive. This observation leads to two challenging tasks: (1) unsupervised accuracy estimation, which aims to estimate the accuracy of a pre-trained classifier on a set of unlabeled test inputs; (2) error detection, which aims to identify mis-classified test inputs. In this paper, we propose a principled and practically effective framework that simultaneously addresses the two tasks.

OpenAI's decision to replace 4o with the more straightforward GPT-5 follows a steady drumbeat of news about the potentially harmful effects of extensive chatbot use. Reports of incidents in which ChatGPT sparked psychosis in users have been everywhere for the past few months, and in a blog post last week, OpenAI acknowledged 4o's failure to recognize when users were experiencing delusions. The company's internal evaluations indicate that GPT-5 blindly affirms users much less than 4o did. AI companionship is new, and there's still a great deal of uncertainty about how it affects people. Yet the experts we consulted warned that while emotionally intense relationships with large language models may or may not be harmful, ripping those models away with no warning almost certainly is.

In the original approach proposed by Devlin et al. [2018], a fraction of selected masked words is

The CSGM framework (Bora-Jalal-Price-Dimakis'17) has shown that deep generative priors can be powerful tools for solving inverse problems.

In this section, we give an overview of related work in stable neural ODE networks. We also give an overview of common adversarial attacks and recent works that defend against adversarial examples. Stable Neural Network Gradient vanishing and gradient exploding are two well-known phenomena in deep learning [1]. The gradient of the objective function, which strongly relies on the training method as well as the neural network architecture, indicates how sensitive the output is with respect to (w.r.t.) input perturbation. Exploding gradient implies instability of the output w.r.t. the input and thus resulting in a non-robust learning architecture.

Graph contrastive learning (GCL), leveraging graph augmentations to convert graphs into different views and further train graph neural networks (GNNs), has achieved considerable success on graph benchmark datasets. Y et, there are still some gaps in directly applying existing GCL methods to real-world data. First, handcrafted graph augmentations require trials and errors, but still can not yield consistent performance on multiple tasks.

ABL introduces a two-stage gradient ascent mechanism for standard training to 1) help isolate backdoor examples at an early training stage, and 2) break the correlation between backdoor examples and the target class at a later training stage.

Efficiently approximating local curvature information of the loss function is a key tool for optimization and compression of deep neural networks.