Conflict-Driven Satisfiability for Theory Combination: Lemmas, Modules, and Proofs

International audience

Generative AI has shown its value for many software engineering tasks. Still in its infancy, large language model (LLM)-based proof generation lags behind LLM-based code generation. In this paper, we present AutoVerus. AutoVerus uses LLMs to automatically generate correctness proof for Rust code. AutoVerus is designed to match the unique features of Verus, a verification tool that can prove the correctness of Rust code using proofs and specifications also written in Rust. AutoVerus consists of a network of agents that are crafted and orchestrated to mimic human experts' three phases of proof construction: preliminary proof generation, proof refinement guided by generic tips, and proof debugging guided by verification errors. To thoroughly evaluate AutoVerus and help foster future research in this direction, we have built a benchmark suite of 150 non-trivial proof tasks, based on existing code-generation benchmarks and verification benchmarks. Our evaluation shows that AutoVerus can automatically generate correct proof for more than 90% of them, with more than half of them tackled in less than 30 seconds or 3 LLM calls.

Mental health professionals have employed a variety of clinical and experimental neuropsychological tests for exploring purported memory alterations in PTSD. Protocols range from standard tests of immediate and delayed learning, recall, and recognition to elaborate paradigms using experimental stimuli for assessment of information-processing skills. Whereas the former have typically focused on general learning and memory capabilities, experimental paradigms have examined the role of trauma-related cues and their impact on remembering. Findings to date suggest that memory abilities in PTSD patients range from intact to mildly impaired on general tests of verbal or visual memory. At the same time, memory tests involving trauma-specific stimuli point to alterations in cognitive information processing, specifically, an attentional bias manifested by changes in speed, accuracy, and depth of processing. The role of a semantic information network involving enhanced specificity for trauma cues is discussed along with possible implications for brain structures and theories of PTSD.

Concrete is one of the most widely used materials in the world, playing a fundamental role in modern human civilization. The analysis of concrete's microstructure contributes to the design and improvement of concrete compositions. Scanning electron microscopy (SEM) imaging is the most common and direct method for analyzing the microstructure of concrete. However, previous research on SEM image analysis has encountered several challenges: low accuracy and efficiency, lack of quantitative research, poor generalization capability, and dependency on labels. The aim of this thesis is to develop a highly automated and accurate method for analyzing concrete SEM images while improving generalization capability through transfer learning. To achieve this, we propose analysis algorithms based on deep learning and design methods to enhance generalization capability and alleviate the reliance on labels. One long-standing problem in the analysis of concrete SEM data is the heavy reliance on researchers' personal experience or pixel-based shallow features, which can only perform simple qualitative analysis and lack automation and precision. In this thesis, we first establish a deep learning-based SEM data classification framework, followed by an investigation into the impact of pre-trained models using transfer learning techniques. The results demonstrate that our approach is more robust and achieves higher accuracy compared to traditional methods. Moreover, further improvements in model performance can be achieved by pre-training on different concrete SEM datasets. The limited efficiency and accuracy of previous algorithms for SEM data analysis have hindered quantitative research in analyzing SEM data. Additionally, the presence of significant noise and irregular boundaries in SEM images also poses considerable challenges for quantitative analysis. To solve these problems, we propose an improved segmentation algorithm, specifically focusing on edge segmentation in SEM images, effectively enhancing the accuracy of SEM image segmentation. We also compare our results with chemical analysis, validating the feasibility of the proposed algorithm. Even with accurate algorithms, the issue of generalization capability cannot be avoided in the analysis of SEM data. When deploying models trained on one SEM dataset to new scenarios, such as different compositions or regions, a significant decrease in performance is often observed. Since the training data and test data belong to different distributions, and directly applying a trained model can be highly risky. Therefore, it is necessary to reduce the differences between data from different sources to ensure algorithm applicability. In this thesis, we propose an algorithm that utilizes an autoencoder structure to selectively exclude training data that deviate from the target data distribution, retaining data that are more representative of the target distribution. This approach enables the training of more generalized models. Results on SEM data demonstrate that our proposed algorithm achieves better performance by removing a portion of the training data. Furthermore, experiments on a popular visual dataset further demonstrate the effectiveness of our method not only for SEM data but also for general tasks. Another generalization capability issue related to models trained on SEM data is the unavailability of labels in new scenarios, which may have data from a different distribution. In such cases, obtaining labels becomes unfeasible. However, traditional research has often overlooked this constraint and instead selected the best results based on the model's performance on new data. To address this challenge, we propose a label-free model evaluation method that combines deep feature information and model parameter information. This method enables the evaluation of different algorithm performances and determines when to stop iteration to avoid overfitting. We conducted extensive experiments on SEM data and popular visual datasets, demonstrating the effectiveness of our method in evaluating the performance of an algorithm, not only for SEM data but also for general data. In summary, this thesis addresses the long-standing issues of low accuracy, low efficiency, poor generalization capability and label dependency in previous studies based on concrete SEM data. We present more accurate and automated algorithms by considering the characteristics of SEM data. Additionally, we enhance generalization capability by proposing an algorithm to reduce the differences between SEM data distributions, significantly improving performance on different domain data. To reduce the reliance on labels, we introduce a label-free evaluation method. These methods are not only effective for the analysis of concrete SEM data but also help with the practical application of general deep learning algorithms.

To explore the impact of blood pressure on cognitive outcomes at 18 years of age in individuals born extremely preterm (<28 weeks' gestation) and at term (≥37 weeks' gestation). Prospective longitudinal cohort comprising 136 young adults born extremely preterm and 120 matched term controls born in Victoria, Australia in 1991 and 1992. Using linear regression, we analysed the relationships between 24-h mean ambulatory blood pressure, systolic and diastolic hypertension with cognitive outcomes. For both birth groups combined, higher 24-h mean ambulatory blood pressure and systolic hypertension were associated with similar or worse cognitive outcomes. The strongest relationships were between higher 24-h mean ambulatory blood pressure and systolic hypertension with poorer general intellect, visual learning and visual memory. We found little evidence that relationships between ambulatory blood pressure and cognitive outcomes differed by birth group. Higher 24-h mean ambulatory blood pressure and systolic hypertension were associated with poorer cognitive outcomes in individuals born extremely preterm and at term, particularly in general intelligence and visual memory.

Perceptual distraction distorts visual working memories. Recent research has shown divergent effects of distraction on memory performance, including attractive biases, impairment of memory precision, and an increase in the guess rate, indicating multiple mechanisms of distraction interference. Here we propose a novel signal-intrusion model based on the TCC (target-confusability-competition) framework to reconcile those discrepant results. We hypothesized that sensory interference is driven by the integration of a target signal and an intrusive distractor signal. Model comparisons showed that this TCC-intrusion model had a superior fit to memory error distributions across three delayed-estimation tasks with distraction (N = 220 adults) compared with other candidate models. According to the model, distractor intrusions decreased along with target-distractor dissimilarity, in accordance with the sensory-recruitment hypothesis. Moreover, TCC-intrusion successfully replicated divergent effects of distraction on memory bias, precision, and guess rate using this one intrusion mechanism. Together, these results suggest that perceptual distractors affect working memories through a unified mechanism of signal intrusion.

Perceptual distraction distorts visual working memories. Recent research has shown divergenteffects of distraction on memory performance, including attractive biases, impairing memoryprecision, and increasing guess rate, indicating multiple mechanisms of distraction interferences.Here, we propose a novel signal intrusion model (TCC-Intrusion) to reconcile those discrepantresults. We hypothesized that sensory interference is driven by the integration of a target signaland an intrusive distractor signal. Model comparisons showed that TCC-Intrusion had a superiorfit to memory error distributions across three delayed-estimation tasks with distraction (N = 220)compared to other candidate models. According to the model, distractor intrusions decreasedalong with target-distractor dissimilarity, in accordance with the sensory recruitment hypothesis.Moreover, TCC-Intrusion successfully replicated divergent effects of distraction on memorybias, precision, and guess using this one intrusion mechanism. Together, these results suggestthat perceptual distractors affect working memories through a unified mechanism of signalintrusion.

Numerous studies with healthy participants indicate that sleep contributes to the consolidation of memory traces by enhancing neuronal plasticity. This sleep-related reconsolidation mechanism is required for the memory representations to become more resistant to interference and being forgotten. However, remarkably little is known of the effect of sleep disorder on different memory processes. For this, here we present four studies to investigate the effect of sleep disorder on different memory processes and consolidation in children and adults. Moreover, our fifth study examined the effect of short-term positive airway pressure treatment in adult patients with obstructive sleep apnea (OSA). In Study I, we investigated the effect of disrupted sleep on declarative and non-declarative forms of learning in children with sleep-disordered breathing (SDB). In Study II, we examined the consolidation of these memory processes in children with SDB. In case of online learning, our results showed that children with SDB exhibited generally weaker declarative memory performance while the online non-declarative learning was preserved. Regarding the offline changes, we found intact consolidation in case of declarative memory as well as sequencespecific and general skill aspects of non-declarative memory in SDB. In Study III, we investigated the more attention-demanding working memory performance and less attentiondemanding non-declarative learning in adult OSA patients. In Study IV, we tested the consolidation of general skill and sequence-specific aspects of non-declarative memory. In the case of online learning, we observed that OSA patients showed general skill learning and sequence-specific learning similar to that of controls. In contrast, the working memory performance was impaired in the OSA group. A case of the consolidation on non-declarative learning we revealed differences in offline changes of general skill learning between OSA patients and controls. The control group showed offline improvement from evening to morning, thus they became faster in the morning after the offline period, while the OSA group did not. In contrast, we failed to find differences in the offline changes of sequence-specific knowledge between the groups. Finally, in Study V, we examined the effectiveness of positive airway pressure treatment after two and half month. We revealed significant improvement in the respiratory functions during sleep which led to improvement in sleep pattern and reduced sleepiness. In the case of cognitive functions we observed significant improvement in complex working memory, short-and long-term verbal memory and shortterm visual memory. In contrast, the OSA patients demonstrated significant impairment in long-term visual memory. In the case of anxiety, we found significant improvement in state anxiety level and trend in trait anxiety which was correlated with hypoxic events during sleep. Furthermore, we found a positive correlation between slow wave sleep and executive functions. The respiratory functions and hypoxic events during sleep associated negatively with executive functions and explicit memory.

This study investigates the effect of physics education on students' achievement in a large‐scale quantitative study of pre‐academic high school students throughout the Netherlands. Two aspects of teacher characteristics as perceived by their students are included: their “pleasantness” principally defined by their perceived friendliness and positive feedback and their “centeredness” principally defined by the perceived teacher centeredness in the lessons. Furthermore, this study includes four student aspects: their “general capability,” their “quantity of work,” their “quality of work,” and their “interest in the lessons.” Structural Equation Modeling is used in order to cluster the different variables defining the perceived pleasantness and the perceived centeredness of the teacher and the general capability, interest, and learning attitudes of the students. Furthermore, interrelations among these components and students' achievement are analyzed. Eventually, a very large effect of the students' general capability (61–72%) and a remarkably smaller effect of the remaining parameters (&lt;3%) on achievement are detected. However, one should not yet conclude that teacher effect on high‐achieving‐students' achievement is consistently low. To the contrary, these results should be seen as an incentive to consider nonlinear effects, to vary ones viewpoint and to include more/other variables. In spite of the almost negligible correlation between the measured aspects of the physics teachers and achievement, the correlations between the teacher variables and the remaining student variables are quite significant. Both the perceived pleasantness and the centeredness of the teachers have a significant effect on the interest of their students. Furthermore, the pleasantness of the teacher correlates with the quality of the students' work and the centeredness of the teacher with their quantity of work. © 2012 Wiley Periodicals, Inc. J Res Sci Teach 49: 465–488, 2012

Visual short-term memory (VSTM) enables humans to form a stable and coherent representation of the external world. However, the nature and temporal dynamics of the neural representations in VSTM that support this stability are barely understood. Here we combined human intracranial electroencephalography (iEEG) recordings with analyses using deep neural networks and semantic models to probe the representational format and temporal dynamics of information in VSTM. We found clear evidence that VSTM maintenance occurred in two distinct representational formats which originated from different encoding periods. The first format derived from an early encoding period (250 to 770 ms) corresponded to higher-order visual representations. The second format originated from a late encoding period (1,000 to 1,980 ms) and contained abstract semantic representations. These representational formats were overall stable during maintenance, with no consistent transformation across time. Nevertheless, maintenance of both representational formats showed substantial arrhythmic fluctuations, i.e., waxing and waning in irregular intervals. The increases of the maintained representational formats were specific to the phases of hippocampal low-frequency activity. Our results demonstrate that human VSTM simultaneously maintains representations at different levels of processing, from higher-order visual information to abstract semantic representations, which are stably maintained via coupling to hippocampal low-frequency activity.

The power of state-of-the-art automated and interactive theorem provers has reached the level at which a significant portion of non-trivial mathematical contents can be formalized almost fully automatically. In this paper we present our framework for the formalization of mathematical knowledge that can produce machine verifiable proofs (for different proof assistants) but also human-readable (nearly textbook-like) proofs. As a case study, we focus on one of the twentieth century classics – a book on Tarski’s geometry. We tried to automatically generate such proofs for the theorems from this book using resolution theorem provers and a coherent logic theorem prover. In the first experiment, we used only theorems from the book, in the second we used additional lemmas from the existing Coq formalization of the book, and in the third we used specific dependency lists from the Coq formalization for each theorem. The results show that 37 % of the theorems from the book can be automatically proven (with readable and machine verifiable proofs generated) without any guidance, and with additional lemmas this percentage rises to 42 %. These results give hope that the described framework and other forms of automation can significantly aid mathematicians in developing formal and informal mathematical knowledge.

In recent decades, cyber-physical systems developed using Model-Driven Engineering (MDE) techniques have become ubiquitous in safety-critical domains. Safety assurance cases (ACs) are structured arguments designed to comprehensively show that such systems are safe; however, the reasoning steps, or strategies, used in AC arguments are often informal and difficult to rigorously evaluate. Consequently, AC arguments are prone to fallacies, and unsafe systems have been deployed as a result of fallacious ACs. To mitigate this problem, prior work [32] created a set of provably valid AC strategy templates to guide developers in building rigorous ACs. Yet instantiations of these templates remain error-prone and still need to be reviewed manually. In this paper, we report on using the interactive theorem prover Lean to bridge the gap between safety arguments and rigorous model-based reasoning. We generate formal, modelbased machine-checked AC arguments, taking advantage of the traceability between model and safety artifacts, and mitigating errors that could arise from manual argument assessment. The approach is implemented in an extended version of the MMINT-A model management tool [10]. Implementation includes a conversion of informal claims into formal Lean properties, decomposition into formal sub-properties and generation of correctness proofs. We demonstrate the applicability of the approach on two safety case studies from the literature.

Posttraumatic Stress Disorder: Memory and Learning Performance in Children and Adolescents

Fetal alcohol syndrome (FAS) is the most common nonhereditary cause of mental retardation, with deficits in general intellectual functioning, learning, memory, attention, and problem-solving. Presented here is the first case in which measured neurocognitive abilities were determined before, during, and after hyperbaric oxygen therapy in a case of FAS involving a teenage male patient. Memory, reaction time, and visual motor speed assessments were compared. After 40 hyperbaric treatments with 100% oxygen at 1.5 atmospheres absolute, the patient's performance in 6 of 6 categories of the computer-administered test battery improved. Word composite (verbal) scores improved from 55% to 73%, memory composite (visual) scores improved from 38% to 55%, reaction time composites improved from 1.03 to 0.53 seconds, impulse control composite scores improved from 8 to 5, and visual motor speed scores improved from 18.6 to 19.03. The patient's subjective symptoms diminished 94%. Six months after these treatments, the patient's verbal memory was maintained at 73% without any other interventions; impulsivity continued to improve, whereas other indices did not. Thirty-three additional treatments continued to improve test performance, with verbal memory at 95%, visual memory at 57%, and a 100% reduction of subjective symptoms. This patient, with 15-year-matured FAS, benefited from a short course of low-pressure hyperbaric oxygen therapy, sustained durable cognitive improvements, and continued to exhibit improvement with another short course of treatments.

Abstract Factor analyses of clinical memory tests have demonstrated that scores on tests purported to measure "visual memory" load highly on visuoperceptual and visuoconstructional factors with modest loadings on measures that define general learning and memory. This may reflect the multifactorial nature of visual-spatial memory and a multitude of methodological and test construction issues. This paper is intended to provide the clinical neuropsychologist with an explanation as to why the search for a pure measure of visual memory has been an arduous one and identifies five reasons for this apparent difficulty. Suggested directions for future nonverbal memory test development are offered.