Reasoning Tasks Research Articles

Domain Ontology-Driven Knowledge Graph Generation from Text

A knowledge graph serves as a unified and standardized representation for extracting and representing textual information. In the field of knowledge extraction and representation research, named entity recognition and relation extraction provide effective solutions for knowledge graph generation tasks. However, it is a challenge that lies in extracting domain-specific knowledge from the rich and general textual corpora and generating corresponding domain knowledge graphs to support domain-specific reasoning, question-answering, and decision-making tasks. The hierarchical domain knowledge representation model (i.e. domain ontology) provides a solution for this problem. Therefore, we propose an end-to-end approach based on domain ontology embedding and pre-trained language models for domain knowledge graph generation from text, which incorporates domain node recognition and domain relation extraction phases. We evaluated our domain ontology-driven model on the Wikidata-TekGen dataset and the DBpedia-WebNLG dataset, and the results indicate that our approach based on the pre-trained language models with fewer parameters compared with the baseline models has significantly contributed to the domain knowledge graph generation without prompts.

Neural Reasoning Machine for Recommendation

Most of the existing recommendation models are designed based on the principles of learning and matching: by learning the user and item embeddings and using learned or designed functions as matching models, they try to explore the similarity pattern between users and items for recommendation. However, recommendation is not only a perceptual matching task, but also a cognitive reasoning task because user behaviors are not merely based on item similarity but also based on users’ careful reasoning about what they need and what they want.

Proof Theory and Decision Procedures for Deontic STIT Logics

This paper provides a set of cut-free complete sequent-style calculi for deontic STIT (‘See To It That’) logics used to formally reason about choice-making, obligations, and norms in a multi-agent setting. We leverage these calculi to write a proof-search algorithm deciding deontic, multi-agent STIT logics with (un)limited choice and introduce a loop-checking mechanism to ensure the termination of the algorithm. Despite the acknowledged potential for deontic reasoning in the context of autonomous, multi-agent scenarios, this work is the first to provide a syntactic decision procedure for this class of logics. Our proofsearch procedure is designed to provide verifiable witnesses/certificates of the (in)validity of formulae, which permits an analysis of the (non)theoremhood of formulae and act as explanations thereof. We show how the proof system and decision algorithm can be used to automate normative reasoning tasks such as duty checking (viz. determining an agent’s obligations relative to a given knowledge base), compliance checking (viz. determining if a choice, considered by an agent as potential conduct, complies with the given knowledge base), and joint fulfillment checking (viz. determining whether under a specified factual context an agent can jointly fulfill all their duties).

How does teaching experience impact brain processes underlying the theory of mind? Study on primary school educators

ABSTRACT Despite its importance for daily social interactions, few studies have explored interindividual differences in the Theory of Mind (ToM) abilities of healthy adults. We used Children’s False-Attribution (CFA), Children’s False-Beliefs (CFB), and Belief-Desire Reasoning tasks, along with fMRI-based assessments, in a comparative analysis of ToM among primary school teachers (PST; n = 27), skilled in social interactions with children, and matched controls (MC; n = 24), who lacked such experience. PST demonstrated slower reaction times than MC in Adult and Child false-belief stories of CFB. However, no other behavioral differences between the groups and between-group differences were observed at the brain level. Both groups presented similar valence ratings for stories in the CFA. Notably, fMRI analysis revealed a group-by-condition interaction effect in the right lateral orbitofrontal cortex (OFC). In PSTs, OFC activation decreased during negative false-attribution stories regarding children compared to stories related to adults, whereas MC demonstrated an opposite activation pattern. Between-group differences in right lateral OFC activity possibly signify a neural efficiency effect secondary to frequent social interactions of PSTs, unlike the MCs, with children in the working environment. These results underscore the significance of everyday social experiences in the functional plasticity of ToM networks.

Integrating Neural-Symbolic Reasoning With Variational Causal Inference Network for Explanatory Visual Question Answering.

Recently, a novel multimodal reasoning task named Explanatory Visual Question Answering (EVQA) has been introduced, which combines answering visual questions with multimodal explanation generation to expound upon the underlying reasoning processes. In contrast to conventional Visual Question Answering (VQA) that merely concentrates on providing answers, EVQA aims to improve the explainability and verifiability of reasoning by providing user-friendly explanations. Despite the improved explainability of inferred results, the existing EVQA models still adopt black-box neural networks to infer results, lacking the explainability of the reasoning process. Moreover, existing EVQA models commonly predict answers and explanations in isolation, overlooking the inherent causal correlation between them. To handle these challenges, we propose a Program-guided Variational Causal Inference Network (Pro-VCIN) that integrates neural-symbolic reasoning with variational causal inference and constructs causal correlations between the predicted answers and explanations. First, we utilize pretrained models to extract visual features and convert questions into the corresponding programs. Second, we propose a multimodal program Transformer to translate programs and the related visual features into coherent and rational explanations of the reasoning processes. Finally, we propose a variational causal inference to construct the target structural causal model and predict answers based on the causal correlation to explanations. Comprehensive experiments conducted on EVQA benchmark datasets reveal the superiority of Pro-VCIN in terms of both performance and explainability over state-of-the-art EVQA methods.

An Investigation of the Associations Between Counterfactual Reasoning and Executive Functions in Chinese Preschool Children

ABSTRACT Counterfactual reasoning involves reasoning about what could have happened but did not happen in the past. It assists people to learn from experience to improve future performance. Different from English with subjunctive forms (e.g. if someone had done something), Chinese does not apply subjunctive forms. In Chinese, people often decide that a premise is counterfactual based on their knowledge of the past, which involves working memory. This study investigated counterfactual reasoning performance and the associations between counterfactual reasoning and executive functions in 257 Chinese 4- to 6-year-olds. Children completed counterfactual reasoning, working memory, inhibitory control, and receptive language tasks. Results showed that Chinese 4-year-olds achieved above-chance counterfactual reasoning performance. Also, working memory was significantly associated with counterfactual reasoning. Hence, Chinese children might solve counterfactual reasoning tasks without complex subjunctive forms relatively early, and working memory might assist counterfactual reasoning in Chinese.

Exploring spatial reasoning performances of CNN on linear layout dataset

Abstract Spatial reasoning, a fundamental aspect of human intelligence, is essential for machine learning models to understand and interpret object relationships. It is crucial for numerous real-world applications, ranging from autonomous navigation to urban planning. The lack of comprehensive datasets limits the development and evaluation of models that can effectively handle spatial reasoning tasks. Existing datasets often contain complex spatial reasoning problems with overlapping spatial relationships, making it challenging to diagnose specific aspects that a model struggles with. We address this gap by introducing a new dataset of linear layouts. This dataset is systematically designed to exhibit a range of spatial relations and complexity levels. Analyzing spatial reasoning through linear layout generation offers a more structured and manageable approach to understanding how models learn and interpret spatial relationships. Linear layout generation has broad applicability and is of fundamental importance in design and optimization. To benchmark dataset, we develop LinLayCNN, a generic data-driven method that applies shallow, one-dimensional convolutional neural network (CNN), to generate linear layouts in an iterative process. Experimental results reveal that LinLayCNN can effectively solve fundamental spatial challenges even with the relatively small size of the training set. It is capable of precise object placement, making it a robust tool for linear layout generation. Current layout generation methods focus on domain-specific solutions, often fail to maintain the precision needed for technical domains, such as accurate sizing, and object counting. They also require a substantial amount of data to function effectively. LinLayCNN overcame these issues. This study also enhances our understanding of CNNs' capabilities in spatial reasoning, highlight their potential to advance the field of layout generation. As a result, our approach establishes a clear benchmark for evaluating spatial reasoning and aids in development of models that can more effectively understand and reason about space.

Reasoning in social versus non-social domains and its relation to autistic traits.

Enhanced rationality has been linked to higher levels of autistic traits, characterised by increased deliberation and decreased intuition, alongside reduced susceptibility to common reasoning biases. However, it is unclear whether this is domain-specific or domain-general. We aimed to explore whether reasoning tendencies differ across social and non-social domains in relation to autistic traits. We conducted two experiments (N1 = 72, N2 = 217) using a reasoning task with social and non-social scenario comparisons to evaluate the specific information participants used when making judgments about children, in the social domain, and objects, in the non-social domain. We consistently found a greater reliance on behaviour-based information in the non-social domain, compared to the social domain, indicating a more deliberative approach. In Experiment 1, we found a correlation between autistic traits and the proportion of behaviour-based information, suggesting a more deliberative approach, when making judgments about children, and not about objects. In Experiment 2, with a larger sample, shortened version of the reasoning task, and requests for written justification, we did not identify a significant correlation between these variables. With this study, we introduce a novel scenario-based reasoning task that systematically compares the social and non-social domains. Our findings highlight the complex nature of the relationship between reasoning style and autistic traits.

Neural networks for abstraction and reasoning

For half a century, artificial intelligence research has attempted to reproduce the human qualities of abstraction and reasoning - creating computer systems that can learn new concepts from a minimal set of examples, in settings where humans find this easy. While specific neural networks are able to solve an impressive range of problems, broad generalisation to situations outside their training data has proved elusive. In this work, we look at several novel approaches for solving the ion & Reasoning Corpus (ARC). This is a dataset of abstract visual reasoning tasks introduced to test algorithms on broad generalization. Despite three international competitions with $100,000 in prizes, the best algorithms still fail to solve a majority of ARC tasks. The best solvers today rely on complex hand-crafted rules, without using machine learning at all. We revisit whether recent advances in neural networks allow progress on this task, or whether an entirely different class of models are required. First, we adapt the DreamCoder neurosymbolic reasoning solver to ARC. DreamCoder automatically writes programs in a bespoke domain-specific language to perform reasoning, using a neural network to mimic human intuition. We present the Perceptual ion and Reasoning Language (PeARL) language, which allows DreamCoder to solve ARC tasks, and propose a new recognition model that allows us to significantly improve on the previous best implementation. We also propose a new encoding and augmentation scheme that allows large language models (LLMs) to solve ARC tasks, and find that the largest models can solve some ARC tasks. LLMs are able to solve a different group of problems to state-of-the-art solvers, and provide an interesting way to complement other approaches. We perform an ensemble analysis, combining systems to achieve better results than any system alone and analysing individual strengths. However, it is sobering to see that approaches based on neural networks still lag behind existing hand-crafted solvers, and we suggest avenues for future improvements. Our findings with the ensemble model may indicate that a diversity of methods might be necessary to solve problems in ARC. Humans likely employ diverse strategies to solve ARC. Studies involving human participants to identify the strategies they employ to solve ARC could provide valuable insights for future AI approaches. Finally, we publish the arckit Python library to make future research on ARC easier.

Cause and Effect: Can Large Language Models Truly Understand Causality?

With the rise of Large Language Models (LLMs), it has become crucial to understand their capabilities and limitations in deciphering and explaining the complex web of causal relationships that language entails. Current methods use either explicit or implicit causal reasoning, yet there is a strong need for a unified approach combining both to tackle a wide array of causal relationships more effectively. This research proposes a novel architecture called Context-Aware Reasoning Enhancement with Counterfactual Analysis (CARE-CA) to enhance causal reasoning and explainability. The proposed framework incorporates an explicit causal detection module with ConceptNet and counterfactual statements, as well as implicit causal detection through LLMs. Our framework goes one step further with a layer of counterfactual explanations to accentuate LLMs' understanding of causality. The knowledge from ConceptNet enhances the performance of multiple causal reasoning tasks such as causal discovery, causal identification, and counterfactual reasoning. The counterfactual sentences add explicit knowledge of `not caused by' scenarios. By combining these powerful modules, our model aims to provide a deeper understanding of causal relationships, enabling enhanced interpretability. Evaluation of benchmark datasets shows improved performance across all metrics, such as accuracy, precision, recall, and F1 scores. We also present CausalNet, a novel dataset specifically curated to benchmark and enhance the causal reasoning capabilities of LLMs. This dataset is accompanied by code designed to facilitate further research in this domain.

SecKG2vec: A novel security knowledge graph relational reasoning method based on semantic and structural fusion embedding

Knowledge graph technology is widely used in network security design, analysis, and detection. By collecting, organizing, and mining various security knowledge, it provides scientific support for security decisions. Some public Security Knowledge Repositories (SKRs) are frequently used to construct security knowledge graphs. The quality of SKRs affects the efficiency and effectiveness of security analysis. However, the current situation is that the identification of relational information among security knowledge elements is not sufficient and timely, and a large number of key relational information is missing. In view of this, we propose a security knowledge graph relational reasoning method, based on the fusion embedding of semantic correlation and structure correlation, named SecKG2vec. By SecKG2vec, the embedded vector simultaneously presents both semantic and structural characteristics, and it can exhibit better relational reasoning performance. In qualitative evaluation and quantitative experiments with baseline methods, SecKG2vec has better performance in relationship reasoning task and entity reasoning task, and potential capability of 0-shot scenario prediction.

Evaluating the Performance of ChatGPT in the Prescribing Safety Assessment: Implications for Artificial Intelligence-Assisted Prescribing.

Objective With the rapid advancement of artificial intelligence (AI) technologies, models likeChat Generative Pre-Trained Transformer (ChatGPT) are increasingly being evaluated for their potential applications in healthcare. The Prescribing Safety Assessment (PSA) is a standardised test for junior physicians in the UK to evaluate prescribing competence. This study aims to assess ChatGPT's ability to pass the PSA and its performance across different exam sections. Methodology ChatGPT (version GPT-4) was tested on four official PSA practice papers, each containing 30 questions, in three independent trials per paper, with answers evaluated using official PSA mark schemes. Performance was measured by calculating overall percentage scores and comparing them to the pass marks provided for each practice paper. Subsection performance was also analysed to identify strengths and weaknesses. Results ChatGPT achieved mean scores of 257/300 (85.67%), 236/300 (78.67%), 199/300 (66.33%), and 233/300 (77.67%) across the four papers, consistently surpassing the pass marks where available. ChatGPT performed well in sections requiring factual recall, such as "Adverse Drug Reactions", scoring 63/72 (87.50%), and "Communicating Information", scoring 63/72 (88.89%). However, it struggled in "Data Interpretation", scoring 32/72 (44.44%), showing variability across trials and indicating limitations in handling more complex clinical reasoning tasks. Conclusion While ChatGPT demonstrated strong potential in passing the PSA and excelling in sections requiring factual knowledge, its limitations in data interpretation highlight the current gaps in AI's ability to fully replicate human clinical judgement. ChatGPT shows promise in supporting safe prescribing, particularly in areas prone to human error, such as drug interactions and communicating correct information. However, due to its variability in more complex reasoning tasks, ChatGPT is not yet ready to replace human prescribers and should instead serve as a supplemental tool in clinical practice.

In-situ observations of the three-dimensional energy cascade rate and Yaglom flux in the Earth’s magnetosheath

Abstract Measuring the energy cascade rate in space plasmas is a challenging task for several reasons. This quantity is (i) inherently three-dimensional (ii) scale-dependent, (iii) anisotropic in the interplanetary plasma, and (iv) requires measurements of plasma parameters in at least four points. Here, it is shown how three of these problems have been addressed by applying the novel lag polyhedra derivative ensemble (LPDE) technique to the magnetospheric multiscale (MMS) mission in the Earth’s magnetosheath. In particular, this technique solves the full vectorial Yaglom equation, handling problem (i), does not require the assumption of isotropy, solving problem (iii), while the application to MMS addresses constraint (iv).

Two-stage Rule-induction visual reasoning on RPMs with an application to video prediction

Raven’s Progressive Matrices (RPMs) are frequently used in evaluating human’s visual reasoning ability. Researchers have made considerable efforts in developing systems to automatically solve the RPM problem, often through a black-box end-to-end convolutional neural network for both visual recognition and logical reasoning tasks. Based on the intrinsic natures of RPM problem, we propose a Two-stage Rule-Induction Visual Reasoner (TRIVR), which consists of a perception module and a reasoning module, to tackle the challenges of real-world visual recognition and subsequent logical reasoning tasks, respectively. For the reasoning module, we further propose a “2+1” formulation that models human’s thinking in solving RPMs and significantly reduces the model complexity. It derives a reasoning rule from each RPM sample, which is not feasible for existing methods. As a result, the proposed reasoning module is capable of yielding a set of reasoning rules modeling human in solving the RPM problems. To validate the proposed method on real-world applications, an RPM-like Video Prediction (RVP) dataset is constructed, where visual reasoning is conducted on RPMs constructed using real-world video frames. Experimental results on various RPM-like datasets demonstrate that the proposed TRIVR achieves a significant and consistent performance gain compared with state-of-the-art models.

L2 Learners’ Perspectives on Data-Driven Learning for Identifying Properties of Near-Synonymous Words: A Convergent Mixed-Methods Study

This study examines second language (L2) learners’ perspectives regarding the affordances and challenges of using the Data-Driven Learning (DDL) to identify the properties of near-synonymous words. Employing a convergent mixed-method design, this study deciphers the perceptions of 40 undergraduate L2 learners majoring in English language teaching. After an initial identification of the learners’ vocabulary levels, the experienced benefits and barriers associated with carrying out experiential tasks were elicited via questionnaire data and open-ended survey questions. Descriptive statistics, including means and standard deviations, were revealed and thematic analyses of the responses to the survey questions were documented. The results indicate that completing tasks through the corpus was found to enhance their knowledge of collocations. Integrating corpus tasks into YouGlish (an online practice tool for authentic spoken English in context) was found to increase their awareness of the contextual properties of words. The identification of condensed language exposure, lexical inference, and elicitation of flexible and context-specific patterns were reported to be beneficial. Acknowledging these benefits, gaining familiarity with the corpus interface, encountering limited access to search queries, and analyzing large amounts of concordance lines posed challenges for learners. This research presents the implementation of the DDL supported by experiential learning, contextually rich input, and inductive reasoning tasks in vocabulary learning by further offering instructional implications in L2 contexts.

Learning differentiable logic programs for abstract visual reasoning

AbstractVisual reasoning is essential for building intelligent agents that understand the world and perform problem-solving beyond perception. Differentiable forward reasoning has been developed to integrate reasoning with gradient-based machine learning paradigms. However, due to the memory intensity, most existing approaches do not bring the best of the expressivity of first-order logic, excluding a crucial ability to solve abstract visual reasoning, where agents need to perform reasoning by using analogies on abstract concepts in different scenarios. To overcome this problem, we propose NEUro-symbolic Message-pAssiNg reasoNer (NEUMANN), which is a graph-based differentiable forward reasoner, passing messages in a memory-efficient manner and handling structured programs with functors. Moreover, we propose a computationally-efficient structure learning algorithm to perform explanatory program induction on complex visual scenes. To evaluate, in addition to conventional visual reasoning tasks, we propose a new task, visual reasoning behind-the-scenes, where agents need to learn abstract programs and then answer queries by imagining scenes that are not observed. We empirically demonstrate that NEUMANN solves visual reasoning tasks efficiently, outperforming neural, symbolic, and neuro-symbolic baselines.

Thalamus and its functional connections with cortical regions contribute to complexity-dependent cognitive reasoning

The thalamus is crucial for supporting various cognitive behaviors due to its extensive connectivity with multiple cortical regions. However, the role of the thalamus and its functional connections with cortical regions in cognitive reasoning remains unclear, since previous research has mainly focused on cortical regions when studying the neural mechanisms underlying cognitive reasoning. To fill this knowledge gap, we utilized 7 T functional magnetic resonance imaging (fMRI) to study the activation patterns of the thalamus and its functional connections with cortical regions during cognitive reasoning task, while also examining how the complexity of reasoning tasks affects thalamic activation and functional connections with cortical regions. Our findings showed that cognitive reasoning processes are related to increased activation of the thalamus and its functional connections with a specific set of cortical regions, consisting of dorsolateral prefrontal cortex, inferior frontal sulcus, intraparietal sulcus, anterior cingulate cortex/presupplementary motor area, precuneus, and ventral medial prefrontal cortex. Moreover, the increase in relational complexity of the reasoning tasks led to a corresponding increase in thalamic activation and functional connectivity with cortical regions. Given the complex thalamus structure, including multiple distinct nuclei exhibiting specific functional connections with particular cortical regions, we used an atlas defined thalamic subdivisions based on its structural connectivity with different cortical regions. Our findings indicated that these different thalamic subregions not only exhibited distinct connectivity patterns with specific cortical regions during performance of cognitive reasoning, but also showed distinct connectivity patterns varied with task complexity. Overall, our study presents evidence of the thalamus’s role and its connections with cortical regions in supporting increasingly complex cognitive reasoning behavior, illuminating its contribution to higher-order cognitive functions, such as reasoning.

Distributional hypothesis as isomorphism between word-word co-occurrence and analogical parallelograms.

Most of the modern natural language processing (NLP) techniques are based on the vector space models of language, in which each word is represented by a vector in a high dimensional space. One of the earliest successes was demonstrated by the four-term analogical reasoning task: what is to C as B is to A? The trained word vectors form "parallelograms" representing the quadruple of words in analogy. This discovery in NLP offers us insight into our understanding of human semantic representation of words via analogical reasoning. Despite successful applications of the large-scale language models, it has not been fully understood why such parallelograms emerge by learning through natural language data. As the vector space model is not optimized to form parallelograms, the key structure related to geometric shapes of word vectors is expected to be in the data, rather than the models. In the present article, we test our hypothesis that such parallelogram arrangement of word vectors readily exists in the co-occurrence statistics of language. Our approach focuses more on the data itself, and it is different from the existing theoretical approach trying to find the mechanism of parallelogram formation in the algorithms and/or vector arithmetic operations on word vectors. First, our analysis suggested that analogical reasoning is possible by decomposition of the bigram co-occurrence matrix. Second, we demonstrated the formation of a parallelepiped, a more structured geometric object than a parallelogram, by creating a small artificial corpus and its word vectors. With these results, we propose a refined form of distributional hypothesis pointing out an isomorphism between a sort of symmetry or exchangeability and word co-occurrence statistics.

Social cognition in adults with neurofibromatosis type 1.

Adult patients with the genetic disease neurofibromatosis type 1 (NF1) frequently report social difficulties. To date, however, only two studies have explored whether these difficulties are caused by social cognition deficits, and these yielded contradictory data. The aim of the present study was to exhaustively assess social cognition abilities (emotion, theory of mind, moral reasoning, and social information processing) in adults with NF1, compared with a control group, and to explore links between social cognition and disease characteristics (mode of inheritance, severity, and visibility). We administered a social cognition battery to 20 adults with NF1 (mean age = 26.5 years, SD = 7.4) and 20 healthy adults matched for sociodemographic variables. Patients scored significantly lower than controls on emotion, theory of mind, moral reasoning, and social information processing tasks. No effects of disease characteristics were found. These results appear to confirm that adults with NF1 have a social cognition weaknesses that could explain, at least in part, their social difficulties, although social abilities are not all impaired to the same extent. Regarding the impact of the disease characteristics, the patient sample seemed slightly insufficient for the power analyses performed. Thus, this exploratory study should form the basis of further research, with the objective of replicating these results with larger and more appropriately matched samples.

CHA: Conditional Hyper-Adapter method for detecting human–object interaction

Human–object interactions (HOI) detection aims at capturing human–object pairs in images and predicting their actions. It is an essential step for many visual reasoning tasks, such as VQA, image retrieval and surveillance event detection. The challenge of this task is to tackle the compositional learning problem, especially in a few-shot setting. A straightforward approach is designing a group of dedicated models for each specific pair. However, the maintenance of these independent models is unrealistic due to combinatorial explosion. To address the above problems, we propose a new Conditional Hyper-Adapter (CHA) method based on meta-learning. Different from previous works, our approach regards each <verb, object> as an independent sub-task. Meanwhile, we design two kinds of Hyper-Adapter structures to guide the model to learn “how to address the HOI detection”. By combining the different conditions and hypernetwork, the CHA can adaptively generate partial parameters and improve the representation and generalization ability of the model. Finally, our proposed method can be viewed as a plug-and-play module to boost existing HOI detection models on the widely used HOI benchmarks.