In recent years, proper scoring rules have emerged as a power ful general approach for estimating probability distributions. In addition to significantly ex panding the range of modeling techniques that can be applied in practice, this has also substantially broadened the conceptual understanding of estimation methods. Originally, proper scoring rules we re conceived in meteorology as summary statistics for describing the performance of probabilisti c forecasts ( Murphy and Winkler, 1984), but they also play an important role in economics as tools for bel ief elicitation ( Schotter and Trevino, 2014). A probabilistic forecast is a probability distribution ove r the space of the possible outcomes of the future event that is stated by the forecaster. The simple st and most popular case of probabilistic forecasts arises when the outcome is binary, so the probabilistic forecast reduces to issuing a predictive probability of success. Brier ( 1950) was the first to consider the problem of devising a scoring rule which could not be "played" by a dishonest fore casting agent. He introduced the quadratic scoring rule and showed that it incentivizes a for ecasting agent to state his most accurate probability estimate when faced with uncertainty.

An AI assessor is an external, ideally indepen-dent system that predicts an indicator, e.g., a loss value, of another AI system. Assessors can lever-age information from the test results of many other AI systems and have the flexibility of be-ing trained on any loss function or scoring rule: from squared error to toxicity metrics. Here we address the question: is it always optimal to train the assessor for the target metric? Or could it be better to train for a different metric and then map predictions back to the target metric? Us-ing twenty regression and classification problems with tabular data, we experimentally explore this question for, respectively, regression losses and classification scores with monotonic and non-monotonic mappings and find that, contrary to intuition, optimising for more informative met-rics is not generally better. Surprisingly, some monotonic transformations are promising. For example, the logistic loss is useful for minimis-ing absolute or quadratic errors in regression, and the logarithmic score helps maximise quadratic or spherical scores in classification.

Large Language Models (LLMs) are increasingly employed in real-world applications, driving the need to evaluate the trustworthiness of their generated text. To this end, reliable uncertainty estimation is essential. Since current LLMs generate text autoregressively through a stochastic process, the same prompt can lead to varying outputs. Consequently, leading uncertainty estimation methods generate and analyze multiple output sequences to determine the LLM's uncertainty. However, generating output sequences is computationally expensive, making these methods impractical at scale. In this work, we inspect the theoretical foundations of the leading methods and explore new directions to enhance their computational efficiency. Building on the framework of proper scoring rules, we find that the negative log-likelihood of the most likely output sequence constitutes a theoretically grounded uncertainty measure. To approximate this alternative measure, we propose G-NLL, which has the advantage of being obtained using only a single output sequence generated by greedy decoding. This makes uncertainty estimation more efficient and straightforward, while preserving theoretical rigor. Empirical results demonstrate that G-NLL achieves state-of-the-art performance across various LLMs and tasks. Our work lays the foundation for efficient and reliable uncertainty estimation in natural language generation, challenging the necessity of more computationally involved methods currently leading the field.

Probabilistic regression models trained with maximum likelihood estimation (MLE), can sometimes overestimate variance to an unacceptable degree. This is mostly problematic in the multivariate domain. While univariate models often optimize the popular Continuous Ranked Probability Score (CRPS), in the multivariate domain, no such alternative to MLE has yet been widely accepted. The Energy Score - the most investigated alternative - notoriously lacks closed-form expressions and sensitivity to the correlation between target variables. In this paper, we propose Conditional CRPS: a multivariate strictly proper scoring rule that extends CRPS. We show that closed-form expressions exist for popular distributions and illustrate their sensitivity to correlation. We then show in a variety of experiments on both synthetic and real data, that Conditional CRPS often outperforms MLE, and produces results comparable to state-of-the-art non-parametric models, such as Distributional Random Forest (DRF).

Language generation based on maximum likelihood estimation (MLE) has become the fundamental approach for text generation. Maximum likelihood estimation is typically performed by minimizing the log-likelihood loss, also known as the logarithmic score in statistical decision theory. The logarithmic score is strictly proper in the sense that it encourages honest forecasts, where the expected score is maximized only when the model reports true probabilities. Although many strictly proper scoring rules exist, the logarithmic score is the only local scoring rule among them that depends exclusively on the probability of the observed sample, making it capable of handling the exponentially large sample space of natural text. In this work, we propose a straightforward strategy for adapting scoring rules to language generation, allowing for language modeling with any non-local scoring rules. Leveraging this strategy, we train language generation models using two classic strictly proper scoring rules, the Brier score and the Spherical score, as alternatives to the logarithmic score. Experimental results indicate that simply substituting the loss function, without adjusting other hyperparameters, can yield substantial improvements in model's generation capabilities. Moreover, these improvements can scale up to large language models (LLMs) such as LLaMA-7B and LLaMA-13B. Source code: \url{https://github.com/shaochenze/ScoringRulesLM}.

Within the scope of natural language processing, the domain of multi-label text classification is uniquely challenging due to its expansive and uneven label distribution. The complexity deepens due to the demand for an extensive set of annotated data for training an advanced deep learning model, especially in specialized fields where the labeling task can be labor-intensive and often requires domain-specific knowledge. Addressing these challenges, our study introduces a novel deep active learning strategy, capitalizing on the Beta family of proper scoring rules within the Expected Loss Reduction framework. It computes the expected increase in scores using the Beta Scoring Rules, which are then transformed into sample vector representations. These vector representations guide the diverse selection of informative samples, directly linking this process to the model's expected proper score. Comprehensive evaluations across both synthetic and real datasets reveal our method's capability to often outperform established acquisition techniques in multi-label text classification, presenting encouraging outcomes across various architectural and dataset scenarios.

In the face of uncertainty, predictions ought to quantify their level of confidence (Gneiting and Katzfuss, 2014). This has been recognized for decades in the literature on weather forecasting (Brier, 1950; Murphy, 1977) and probabilistic forecasting (Dawid, 1984; Gneiting and Raftery, 2007). Ideally, a prediction specifies a probability distribution over potential outcomes. Such predictions are evaluated and compared by means of proper scoring rules, which quantify their value in a way that rewards truthful prediction (Gneiting and Raftery, 2007). In statistical classification and machine learning, the need for reliable uncertainty quantification has not gone unnoticed, as exemplified by the growing interest in the calibration of probabilistic classifiers (Guo et al., 2017; Vaicenavicius et al., 2019). However, classifier evaluation often focuses on the most likely class (i.e., the mode of the predictive distribution) through the use of classification accuracy and related metrics derived from the confusion matrix (Tharwat, 2020; Hui and Belkin, 2021).

Probability forecasts for binary outcomes, often referred to as probabilistic classifiers or confidence scores, are ubiquitous in science and society, and methods for evaluating and comparing them are in great demand. We propose and study a triptych of diagnostic graphics that focus on distinct and complementary aspects of forecast performance: The reliability diagram addresses calibration, the receiver operating characteristic (ROC) curve diagnoses discrimination ability, and the Murphy diagram visualizes overall predictive performance and value. A Murphy curve shows a forecast's mean elementary scores, including the widely used misclassification rate, and the area under a Murphy curve equals the mean Brier score. For a calibrated forecast, the reliability curve lies on the diagonal, and for competing calibrated forecasts, the ROC and Murphy curves share the same number of crossing points. We invoke the recently developed CORP (Consistent, Optimally binned, Reproducible, and Pool-Adjacent-Violators (PAV) algorithm based) approach to craft reliability diagrams and decompose a mean score into miscalibration (MCB), discrimination (DSC), and uncertainty (UNC) components. Plots of the DSC measure of discrimination ability versus the calibration metric MCB visualize classifier performance across multiple competitors. The proposed tools are illustrated in empirical examples from astrophysics, economics, and social science.

Random forests are an ensemble method relevant for many problems, such as regression or classification. They are popular due to their good predictive performance (compared to, e.g., decision trees) requiring only minimal tuning of hyperparameters. They are built via aggregation of multiple regression trees during training and are usually calculated recursively using hard splitting rules. Recently regression forests have been incorporated into the framework of distributional regression, a nowadays popular regression approach aiming at estimating complete conditional distributions rather than relating the mean of an output variable to input features only - as done classically. This article proposes a new type of a distributional regression tree using a multivariate soft split rule. One great advantage of the soft split is that smooth high-dimensional functions can be estimated with only one tree while the complexity of the function is controlled adaptive by information criteria. Moreover, the search for the optimal split variable is obsolete. We show by means of extensive simulation studies that the algorithm has excellent properties and outperforms various benchmark methods, especially in the presence of complex non-linear feature interactions. Finally, we illustrate the usefulness of our approach with an example on probabilistic forecasts for the Sun's activity.

An inductive logic can be formulated in which the elements are not propositions or probability distributions, but information systems. The logic is complete for information systems with binary hypotheses, i.e., it applies to all such systems. It is not complete for information systems with more than two hypotheses, but applies to a subset of such systems. The logic is inductive in that conclusions are more informative than premises. Inferences using the formalism have a strong justification in terms of the expected value of the derived information system.