Although binary classification is a well-studied problem, training reliable classifiers under severe class imbalance remains a challenge. Recent techniques mitigate the ill effects of imbalance on training by modifying the loss functions or optimization methods. We observe that different hyperparameter values on these loss functions perform better at different recall values. We propose to exploit this fact by training one model over a distribution of hyperparameter values-instead of a single value-via Loss Conditional Training (LCT). Experiments show that training over a distribution of hyperparameters not only approximates the performance of several models but actually improves the overall performance of models on both CIFAR and real medical imaging applications, such as melanoma and diabetic retinopathy detection. Furthermore, training models with LCT is more efficient because some hyperparameter tuning can be conducted after training to meet individual needs without needing to retrain from scratch. Consider a classifier that takes images of skin lesions and predicts whether they are melanoma or benign (Rotemberg et al., 2020). Such a system could be especially valuable in underdeveloped countries where expert resources for screening are scarce (Cassidy et al., 2022). The dataset for this problem, along with many other practical problems, is inherently imbalanced (i.e., there are far more benign samples than melanoma samples). Furthermore, there are un-even costs associated with misclassifying the two classes because predicting a benign lesion as melanoma would result in the cost of a biopsy while predicting a melanoma lesion as benign could result in the melanoma spreading before the patient can receive appropriate treatment. Unfortunately, the exact difference in the misclassification costs may not be known a priori and may even change after deployment. For example, the costs may change depending on the amount of biopsy resources available or the prior may change depending on the age and condition of the patient. Thus, a good classifier for this problem should (a) have good performance across a wide range of Precision-Recall tradeoffs and (b) be able to adapt to changes in the prior or misclassification costs.

Compressing high-capability Large Language Models (LLMs) has emerged as a favored strategy for resource-efficient inferences. While state-of-the-art (SoTA) compression methods boast impressive advancements in preserving benign task performance, the potential risks of compression in terms of safety and trustworthiness have been largely neglected. This study conducts the first, thorough evaluation of three (3) leading LLMs using five (5) SoTA compression techniques across eight (8) trustworthiness dimensions. Our experiments highlight the intricate interplay between compression and trustworthiness, revealing some interesting patterns. We find that quantization is currently a more effective approach than pruning in achieving efficiency and trustworthiness simultaneously. For instance, a 4-bit quantized model retains the trustworthiness of its original counterpart, but model pruning significantly degrades trustworthiness, even at 50% sparsity. Moreover, employing quantization within a moderate bit range could unexpectedly improve certain trustworthiness dimensions such as ethics and fairness. Conversely, extreme quantization to very low bit levels (3 bits) tends to reduce trustworthiness significantly. This increased risk cannot be uncovered by looking at benign performance alone, in turn, mandating comprehensive trustworthiness evaluation in practice. These findings culminate in practical recommendations for simultaneously achieving high utility, efficiency, and trustworthiness in LLMs. Code and models are available at https://decoding-comp-trust.github.io.

Although binary classification is a well-studied problem in computer vision, training reliable classifiers under severe class imbalance remains a challenging problem. Recent work has proposed techniques that mitigate the effects of training under imbalance by modifying the loss functions or optimization methods. While this work has led to significant improvements in the overall accuracy in the multi-class case, we observe that slight changes in hyperparameter values of these methods can result in highly variable performance in terms of Receiver Operating Characteristic (ROC) curves on binary problems with severe Figure 1: Distribution of Area Under the ROC Curve (AUC) imbalance. To reduce the sensitivity to hyperparameter values obtained by training the same model on the SIIM-choices and train more general models, ISIC Melanoma classification dataset with 48 different combinations we propose training over a family of loss functions, of hyperparameters on VS loss. Results are shown instead of a single loss function. We develop at three different imbalance ratios. As the imbalance becomes a method for applying Loss Conditional more severe, model performance drops and the Training (LCT) to an imbalanced classification variance in performance drastically increases.

Recent advances in neural image compression (NIC) have produced models that are starting to outperform classic codecs. While this has led to growing excitement about using NIC in real-world applications, the successful adoption of any machine learning system in the wild requires it to generalize (and be robust) to unseen distribution shifts at deployment. Unfortunately, current research lacks comprehensive datasets and informative tools to evaluate and understand NIC performance in real-world settings. To bridge this crucial gap, first, this paper presents a comprehensive benchmark suite to evaluate the out-of-distribution (OOD) performance of image compression methods. Specifically, we provide CLIC-C and Kodak-C by introducing 15 corruptions to the popular CLIC and Kodak benchmarks. Next, we propose spectrally-inspired inspection tools to gain deeper insight into errors introduced by image compression methods as well as their OOD performance. We then carry out a detailed performance comparison of several classic codecs and NIC variants, revealing intriguing findings that challenge our current understanding of the strengths and limitations of NIC. Finally, we corroborate our empirical findings with theoretical analysis, providing an in-depth view of the OOD performance of NIC and its dependence on the spectral properties of the data. Our benchmarks, spectral inspection tools, and findings provide a crucial bridge to the real-world adoption of NIC. We hope that our work will propel future efforts in designing robust and generalizable NIC methods. Code and data will be made available at https://github.com/klieberman/ood_nic.