While generative models, especially large language models (LLMs), are ubiquitous in today's world, principled mechanisms to assess their (in)correctness are limited. Using the conformal prediction framework, previous works construct sets of LLM responses where the probability of including an incorrect response, or error, is capped at a desired user-defined tolerance level. However, since these methods are based on p-values, they are susceptible to p-hacking, i.e., choosing the tolerance level post-hoc can invalidate the guarantees. We therefore leverage e-values to complement generative model outputs with e-scores as a measure of incorrectness. In addition to achieving the same statistical guarantees as before, e-scores provide users flexibility in adaptively choosing tolerance levels after observing the e-scores themselves, by upper bounding a post-hoc notion of error called size distortion. We experimentally demonstrate their efficacy in assessing LLM outputs for different correctness types: mathematical factuality and property constraints satisfaction.

The recent success in language generation capabilities of large language models (LLMs), such as GPT, Bard, Llama etc., can potentially lead to concerns about their possible misuse in inducing mass agitation and communal hatred via generating fake news and spreading misinformation. Traditional means of developing a misinformation ground-truth dataset does not scale well because of the extensive manual effort required to annotate the data. In this paper, we propose an LLM-based approach of creating silver-standard ground-truth datasets for identifying misinformation. Specifically speaking, given a trusted news article, our proposed approach involves prompting LLMs to automatically generate a summarised version of the original article. The prompts in our proposed approach act as a controlling mechanism to generate specific types of factual incorrectness in the generated summaries, e.g., incorrect quantities, false attributions etc. To investigate the usefulness of this dataset, we conduct a set of experiments where we train a range of supervised models for the task of misinformation detection.

Multiple-choice tests are a common approach for assessing candidates' comprehension skills. Standard multiple-choice reading comprehension exams require candidates to select the correct answer option from a discrete set based on a question in relation to a contextual passage. For appropriate assessment, the distractor answer options must by definition be incorrect but plausible and diverse. However, generating good quality distractors satisfying these criteria is a challenging task for content creators. We propose automated assessment metrics for the quality of distractors in multiple-choice reading comprehension tests. Specifically, we define quality in terms of the incorrectness, plausibility and diversity of the distractor options. We assess incorrectness using the classification ability of a binary multiple-choice reading comprehension system. Plausibility is assessed by considering the distractor confidence - the probability mass associated with the distractor options for a standard multi-class multiple-choice reading comprehension system. Diversity is assessed by pairwise comparison of an embedding-based equivalence metric between the distractors of a question. To further validate the plausibility metric we compare against candidate distributions over multiple-choice questions and agreement with a ChatGPT model's interpretation of distractor plausibility and diversity.

When deploying classifiers in the real world, users expect them to respond to inputs appropriately. However, traditional classifiers are not equipped to handle inputs which lie far from the distribution they were trained on. Malicious actors can exploit this defect by making adversarial perturbations designed to cause the classifier to give an incorrect output. Classification-with-rejection methods attempt to solve this problem by allowing networks to refuse to classify an input in which they have low confidence. This works well for strongly adversarial examples, but also leads to the rejection of weakly perturbed images, which intuitively could be correctly classified. To address these issues, we propose Reed-Muller Aggregation Networks (RMAggNet), a classifier inspired by Reed-Muller error-correction codes which can correct and reject inputs. This paper shows that RMAggNet can minimise incorrectness while maintaining good correctness over multiple adversarial attacks at different perturbation budgets by leveraging the ability to correct errors in the classification process. This provides an alternative classification-with-rejection method which can reduce the amount of additional processing in situations where a small number of incorrect classifications are permissible.

Over the last decade, Q&A platforms have played a crucial role in how programmers seek help online. The emergence of ChatGPT, however, is causing a shift in this pattern. Despite ChatGPT's popularity, there hasn't been a thorough investigation into the quality and usability of its responses to software engineering queries. To address this gap, we undertook a comprehensive analysis of ChatGPT's replies to 517 questions from Stack Overflow (SO). We assessed the correctness, consistency, comprehensiveness, and conciseness of these responses. Additionally, we conducted an extensive linguistic analysis and a user study to gain insights into the linguistic and human aspects of ChatGPT's answers. Our examination revealed that 52% of ChatGPT's answers contain inaccuracies and 77% are verbose. Nevertheless, users still prefer ChatGPT's responses 39.34% of the time due to their comprehensiveness and articulate language style. These findings underscore the need for meticulous error correction in ChatGPT while also raising awareness among users about the potential risks associated with seemingly accurate answers.

In this paper, we consider the problem of making distributionally robust, skeptical inferences for the multi-label problem, or more generally for Boolean vectors. By distributionally robust, we mean that we consider a set of possible probability distributions, and by skeptical we understand that we consider as valid only those inferences that are true for every distribution within this set. Such inferences will provide partial predictions whenever the considered set is sufficiently big. We study in particular the Hamming loss case, a common loss function in multi-label problems, showing how skeptical inferences can be made in this setting. Our experimental results are organised in three sections; (1) the first one indicates the gain computational obtained from our theoretical results by using synthetical data sets, (2) the second one indicates that our approaches produce relevant cautiousness on those hard-to-predict instances where its precise counterpart fails, and (3) the last one demonstrates experimentally how our approach copes with imperfect information (generated by a downsampling procedure) better than the partial abstention [31] and the rejection rules.