Natural Question Research Articles

Is (Independent) Subordination Relevant in Equity Derivatives?

ABSTRACTMonroe (1978) demonstrates that any local semimartingale can be represented as a time‐changed Brownian Motion (BM). A natural question arises: does this representation theorem hold when the BM and the time‐change are independent? We prove that a local semimartingale is not equivalent to a BM with a time‐change that is independent from the BM. Our result is obtained utilizing a class of additive processes: the additive normal tempered stable (ATS). This class of processes exhibits an exceptional ability to calibrate the equity volatility surface accurately. We notice that the sub‐class of additive processes that can be obtained with an independent additive subordination is incompatible with market data and shows significantly worse calibration performances than the ATS, especially on short time maturities. These results have been observed every business day in a semester on a dataset of S&P 500 and EURO STOXX 50 options.

What Can Happen When We Look at Art?: An Exploratory Network Model and Latent Profile Analysis of Affective/Cognitive Aspects Underlying Shared, Supraordinate Responses to Museum Visual Art

Art-viewing is a defining component of society and culture, in part because the experience involves a wide-range and nuanced configuration of emotional and cognitive responses. Precisely because of this complexity, however, questions of the actual nature, scope, and variety of art experience remain largely unanswered: what kinds of patterns do we exhibit, how do various components go together, and can these be distilled into shared experiential outcomes? We introduce an exploratory study based on 345 individuals’ unique experiences with one of three sets of artworks. Experiences were assessed via 46 affective and cognitive items based on a recent model, with individuals reporting to what degree they felt each during their encounter. Network and latent profile analyses revealed five patterns, aligning to a Harmonious, Facile, Transformative, and two Negative outcomes. These largely supported model hypotheses, connected to specific appraisals, and could be found, although with varying probability, across individual viewers and artworks.

RGISQL: Integrating Refined Grammatical Information into Relational Graph Neural Network for Text-to-SQL Task

The text-to-SQL task aims to convert natural language questions into corresponding SQL queries based on a given database schema. Previous models that rely on graph neural networks often struggle to accurately capture the complex grammatical relationships present in these questions, leading to poor performance when generating queries for longer requests. To address these challenges, we propose RGISQL, which integrates refined grammatical information extracted from the question and employs segmentation processing to effectively manage long queries. Additionally, RGISQL minimizes the complexity of edge embeddings by reducing the coupling within graph neural networks. By utilizing grammatical dependency trees, RGISQL is better equipped to capture the inherent structure and grammatical rules of questions. This refined grammatical information offers additional contextual and semantic cues for the model, thereby enhancing both its generalizability and interpretability. Furthermore, we dynamically assess the importance of different edges based on the graph structure, which helps reduce the coupling of edge embeddings and further improves the model’s performance. Multiple sets of experiments conducted on the Spider and Spider-Syn datasets demonstrate that RGISQL outperforms other baselines, achieving the best results in both datasets.

Investigating Open Source LLMs to Retrofit Competency Questions in Ontology Engineering

Competency Questions (CQs) are essential in ontology engineering; they express an ontology's functional requirements as natural language questions, offer crucial insights into an ontology's scope and are pivotal for various tasks, e.g. ontology reuse, testing, requirement specification, and pattern definition. Despite their importance, the practice of publishing CQs alongside ontological artefacts is not commonly adopted. We propose an approach based on Generative AI, specifically Large Language Models (LLMs) for retrofitting CQs from existing ontologies and we investigate how open LLMs (i.e. Llama-2-70b, Mistral 7B and Flan-T5-xl) perform in generating CQs for existing ontologies. We compare these results with our previous efforts using closed-source LLMs and we reflect on the results.

The human biological advantage over AI

AbstractRecent advances in AI raise the possibility that AI systems will one day be able to do anything humans can do, only better. If artificial general intelligence (AGI) is achieved, AI systems may be able to understand, reason, problem solve, create, and evolve at a level and speed that humans will increasingly be unable to match, or even understand. These possibilities raise a natural question as to whether AI will eventually become superior to humans, a successor “digital species”, with a rightful claim to assume leadership of the universe. However, a deeper consideration suggests the overlooked differentiator between human beings and AI is not the brain, but the central nervous system (CNS), providing us with an immersive integration with physical reality. It is our CNS that enables us to experience emotion including pain, joy, suffering, and love, and therefore to fully appreciate the consequences of our actions on the world around us. And that emotional understanding of the consequences of our actions is what is required to be able to develop sustainable ethical systems, and so be fully qualified to be the leaders of the universe. A CNS cannot be manufactured or simulated; it must be grown as a biological construct. And so, even the development of consciousness will not be sufficient to make AI systems superior to humans. AI systems may become more capable than humans on almost every measure and transform our society. However, the best foundation for leadership of our universe will always be DNA, not silicon.

Homeostasis control in health and disease by the unfolded protein response.

Cells rely on the endoplasmic reticulum (ER) to fold and assemble newly synthesized transmembrane and secretory proteins - essential for cellular structure-function and for both intracellular and intercellular communication. To ensure the operative fidelity of the ER, eukaryotic cells leverage the unfolded protein response (UPR) - a stress-sensing and signalling network that maintains homeostasis by rebalancing the biosynthetic capacity of the ER according to need. The metazoan UPR can also redirect signalling from cytoprotective adaptation to programmed cell death if homeostasis restoration fails. As such, the UPR benefits multicellular organisms by preserving optimally functioning cells while removing damaged ones. Nevertheless, dysregulation of the UPR can be harmful. In this Review, we discuss the UPR and its regulatory processes as a paradigm in health and disease. We highlight important recent advances in molecular and mechanistic understanding of the UPR that enable greater precision in designing and developing innovative strategies to harness its potential for therapeutic gain. We underscore the rheostatic character of the UPR, its contextual nature and critical open questions for its further elucidation.

PangeBlocks: customized construction of pangenome graphs via maximal blocks

BackgroundThe construction of a pangenome graph is a fundamental task in pangenomics. A natural theoretical question is how to formalize the computational problem of building an optimal pangenome graph, making explicit the underlying optimization criterion and the set of feasible solutions. Current approaches build a pangenome graph with some heuristics, without assuming some explicit optimization criteria. Thus it is unclear how a specific optimization criterion affects the graph topology and downstream analysis, like read mapping and variant calling.ResultsIn this paper, by leveraging the notion of maximal block in a Multiple Sequence Alignment (MSA), we reframe the pangenome graph construction problem as an exact cover problem on blocks called Minimum Weighted Block Cover (MWBC). Then we propose an Integer Linear Programming (ILP) formulation for the MWBC problem that allows us to study the most natural objective functions for building a graph. We provide an implementation of the ILP approach for solving the MWBC and we evaluate it on SARS-CoV-2 complete genomes, showing how different objective functions lead to pangenome graphs that have different properties, hinting that the specific downstream task can drive the graph construction phase.ConclusionWe show that a customized construction of a pangenome graph based on selecting objective functions has a direct impact on the resulting graphs. In particular, our formalization of the MWBC problem, based on finding an optimal subset of blocks covering an MSA, paves the way to novel practical approaches to graph representations of an MSA where the user can guide the construction.

Dynamic multi teacher knowledge distillation for semantic parsing in KBQA

Knowledge base question answering (KBQA) is an important task of extracting answers from a knowledge base by analyzing natural language questions. Semantic parsing methods convert natural language questions into executable logical forms to obtain answers on the knowledge base. Conventional approaches often prioritize singular logical forms, overlooking the distinct strengths inherent in various logical frameworks for problem solving. Recognizing that different logical forms may excel in addressing specific types of questions, our aim is to harness these strengths. By integrating the strengths of different logical forms, we expect to achieve more comprehensive and effective semantic parsing solutions. In our paper, we propose a Dynamic Multi Teacher Knowledge Distillation for Semantic Parsing (DMTKD-SP). DMTKD-SP leverages a collection of teacher models, each mastering a unique logical form, to collaboratively guide a student model so that knowledge from different logical forms can be transferred into the student model. To achieve this, we employ a confidence-based weight assignment module to dynamically assign weights for each teacher model. Furthermore, we introduce a self-distillation mechanism to mitigate the confusion caused by simultaneous learning from multiple teachers. We evaluate DMTKD-SP across variants of the KQA Pro dataset, demonstrating an accuracy improvement of 0.35% on five types of questions, with a notable 0.75% improvement for Count questions.

Robust visual question answering via polarity enhancement and contrast

The Visual Question Answering (VQA) task is an important research direction in the field of artificial intelligence, which requires a model that can simultaneously understand visual images and natural language questions, and answer questions related to images. Recent studies have shown that many Visual Question Answering models rely on statistically regular correlations between questions and answers, which in turn weakens the correlation between visual content and textual information. In this work, we propose an unbiased Visual Question Answering method to solve language priors from the perspective of strengthening the contrast between the correct answer and the positive and negative predictions. We design a new model consisting of two modules with different roles. We input the image and the question corresponding to it into the Answer Visual Attention Modules to generate positive prediction output, and then use a Dual Channels Joint Module to generate negative prediction output with great linguistic prior knowledge. Finally, we input the positive and negative predictions together with the correct answer to our newly designed loss function for training. Our method achieves high performance (61.24%) on the VQA-CP v2 dataset. In addition, most existing debiasing methods improve performance on VQA-CP v2 dataset at the cost of reducing performance on VQA v2 dataset, while our method not only does not reduce the accuracy on VQA v2 dataset. Instead, it improves performance on both datasets mentioned above.

SafeMD: Ownership-Based Safe Memory Deallocation for C Programs

Rust is a relatively new programming language that aims to provide memory safety at compile time. It introduces a novel ownership system that enforces the automatic deallocation of unused resources without using a garbage collector. In light of Rust’s promise of safety, a natural question arises about the possible benefits of exploiting ownership to ensure the memory safety of C programs. In our previous work, we developed a formal ownership checker to verify whether a C program satisfies exclusive ownership constraints. In this paper, we further propose an ownership-based safe memory deallocation approach, named SafeMD, to fix memory leaks in the C programs that satisfy exclusive ownership defined in the prior formal ownership checker. Benefiting from the C programs satisfying exclusive ownership, SafeMD obviates alias and inter-procedural analysis. Also, the patches generated by SafeMD make the input C programs still satisfy exclusive ownership. Usually, a C program that satisfies the exclusive ownership constraints is safer than its normal version. Our evaluation shows that SafeMD is effective in fixing memory leaks of C programs that satisfy exclusive ownership.

Expected 1.x Makespan-Optimal Multi-Agent Path Finding on Grid Graphs in Low Polynomial Time

Multi-Agent Path Finding (MAPF) is NP-hard to solve optimally, even on graphs, suggesting no polynomial-time algorithms can compute exact optimal solutions for them. This raises a natural question: How optimal can polynomial-time algorithms reach? Whereas algorithms for computing constant-factor optimal solutions have been developed, the constant factor is generally very large, limiting their application potential. In this work, among other breakthroughs, we propose the first low-polynomial-time MAPF algorithms delivering 1-1.5 (resp., 1-1.67) asymptotic makespan optimality guarantees for 2D (resp., 3D) grids for random instances at a very high 1/3 agent density, with high probability. Moreover, when regularly distributed obstacles are introduced, our methods experience no performance degradation. These methods generalize to support 100% agent density. Regardless of the dimensionality and density, our high-quality methods are enabled by a unique hierarchical integration of two key building blocks. At the higher level, we apply the labeled Grid Rearrangement Algorithm (GRA), capable of performing efficient reconfiguration on grids through row/column shuffles. At the lower level, we devise novel methods that efficiently simulate row/column shuffles returned by GRA. Our implementations of GRA-based algorithms are highly effective in extensive numerical evaluations, demonstrating excellent scalability compared to other SOTA methods. For example, in 3D settings, GRA-based algorithms readily scale to grids with over 370,000 vertices and over 120,000 agents and consistently achieve conservative makespan optimality approaching 1.5, as predicted by our theoretical analysis.

Statistical methods for clinical trials interrupted by the severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) pandemic: A review.

Cancellation or delay of non-essential medical interventions, limitation of face-to-face assessments or outpatient attendance due to lockdown restrictions, illness or fear of hospital or healthcare centre visits, and halting of research to allow diversion of healthcare resources to focus on the pandemic led to the interruption of many clinical trials during the severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) pandemic. Appropriate analysis approaches are now required for these interrupted trials. In trials with long follow-up and longitudinal outcomes, data may be available on early outcomes for many patients for whom final, primary outcome data were not observed. A natural question is then how these early data can best be used in the trial analysis. Although recommendations are available from regulators, funders, and methodologists, there is a lack of a review of recent work addressing this problem. This article reports a review of recent methods that can be used in the setting of the analysis of interrupted clinical trials with longitudinal outcomes with monotone missingness. A search for methodological papers published during the period 2020-2023 identified 43 relevant publications. We categorised these articles under the four broad themes of missing value imputation, modelling and covariate adjustment, simulation and estimands. Although motivated by the interruption due to SARS-CoV-2 and the resulting disease, the papers reviewed and methods discussed are also relevant to clinical trials interrupted for other reasons, with follow-up discontinued.

Realizing one-dimensional moiré chains with strong electron localization in two-dimensional twisted bilayer WSe2

Two-dimensional (2D) moiré systems based on twisted bilayer graphene and transition metal dichalcogenides provide a promising platform to investigate emergent phenomena driven by strong electron-electron interactions in partially filled flat bands. A natural question arises: Is it possible to expand the 2D correlated moiré physics to one-dimensional (1D) that electron-electron correlation is expected to be further enhanced? This requires selectively doping of 1D moiré chain, which seems to be not within the grasp of today's technology. Therefore, an experimental demonstration of the 1D moiré chain with partially filled electronic states remains absent. Here, we show that we can introduce 1D boundaries, separating two regions with different twist angles, in twisted bilayer WSe2 (tWSe2) by using scanning tunneling microscopy (STM) and demonstrate that the electronic states of 1D moiré sites along the boundaries can be selectively filled. The strong localized charge states of correlated moiré electrons in the 1D moiré chain can be directly imaged and manipulated by combining a back-gate voltage with the STM bias voltage. Our results open the door for realizing new correlated electronic states of the 1D moiré chain in 2D systems.

Harnessing Representative Spatial-Temporal Information for Video Question Answering

Video question answering, aiming to answer a natural language question related to the given video, has become prevalent in the past few years. Although remarkable improvements have been obtained, it is still exposed to the challenge of insufficient comprehension of video content. To this end, we propose a spatial-temporal representative visual exploitation network for video question answering, which enhances the understanding of the video by merely summarizing representative visual information. In order to explore representative object information, we advance adaptive attention based on uncertainty estimation. At the same time, to capture representative frame-level and clip-level visual information, we structure a much more compact set of representations iteratively in an expectation-maximization manner to deprecate noisy information. Both the quantitative and qualitative results on NExT-QA, TGIF-QA, MSRVTT-QA, and MSVD-QA datasets demonstrate the superiority of our model over several state-of-the-art approaches.

Quantum-Inspired Fusion for Open-Domain Question Answering

Open-domain question-answering systems need models capable of referencing multiple passages simultaneously to generate accurate answers. The Rational Fusion-in-Decoder (RFiD) model focuses on differentiating between causal relationships and spurious features by utilizing the encoders of the Fusion-in-Decoder model. However, RFiD reliance on partial token information limits its ability to determine whether the corresponding passage is a rationale for the question, potentially leading to inappropriate answers. To address this issue, we propose a Quantum-Inspired Fusion-in-Decoder (QFiD) model. Our approach introduces a Quantum Fusion Module (QFM) that maps single-dimensional into multi-dimensional hidden states, enabling the model to capture more comprehensive token information. Then, the classical mixture method from quantum information theory is used to fuse all information. Based on the fused information, the model can accurately predict the relationship between the question and passage. Experimental results on two prominent ODQA datasets, Natural Questions and TriviaQA, demonstrate that QFiD outperforms the strong baselines in automatic evaluations.

On Elastic Language Models

Large-scale pretrained language models have achieved compelling performance in a wide range of language understanding and information retrieval tasks. While their large scales ensure capacity, they also hinder deployment. Knowledge distillation offers an opportunity to compress a large language model to a small one, in order to reach a reasonable latency-performance tradeoff. However, for scenarios where the number of requests (e.g., queries submitted to a search engine) is highly variant, the static tradeoff attained by the compressed language model might not always fit. Once a model is assigned with a static tradeoff, it could be inadequate in that the latency is too high when the number of requests is large, or the performance is too low when the number of requests is small. To this end, we propose an elastic language model ( ElasticLM ) that elastically adjusts the tradeoff according to the request stream. The basic idea is to introduce a compute elasticity to the compressed language model, so that the tradeoff could vary on-the-fly along a scalable and controllable compute. Specifically, we impose an elastic structure to equip ElasticLM with compute elasticity and design an elastic optimization method to learn ElasticLM under compute elasticity. To serve ElasticLM , we apply an elastic schedule. Considering the specificity of information retrieval, we adapt ElasticLM to dense retrieval and reranking, and present an ElasticDenser and an ElasticRanker, respectively. Offline evaluation is conducted on a language understanding benchmark GLUE, and several information retrieval tasks including Natural Question, Trivia QA and MS MARCO. The results show that ElasticLM along with ElasticDenser and ElasticRanker can perform correctly and competitively compared with an array of static baselines. Furthermore, an online simulation with concurrency is also carried out. The results demonstrate that ElasticLM can provide elastic tradeoffs with respect to varying request stream.

Optimal circular states

Abstract In quantum optics, superpositions of coherent states, such as Schr"odinger cat states and compass states, and more generally, circular states, have attracted widespread attention due to their nice properties and significant applications. Concerning circular states, a natural question arises as what are the optimal parameters in these states for maximally achieving certain specified quantum features such as average photon number and nonclassicality. It turns out this issue is highly nontrivial and subtle. In this work, we investigate optimal circular states for average photon number, and determine the optimal parameters by a combination of theoretical and numerical methods. In particular, we establish several analytical results and also some rather detailed numerical results. We tabulate some numerical results, which may be useful in both theoretical and experimental studies of superpositions of coherent states.

The Three Postulate Idea: Between Limits and Metaphysical Beliefs in Immanuel Kant's System of Philosophy

In the history of Western philosophy, debates about truth and error have continued to develop, reflecting diverse views that have emerged over time. The question of human nature became the center of intellectual attention, expanding philosophical study from speculation to fundamental thinking. Amid this context, Immanuel Kant emerged as a key figure who brought a new approach to philosophical inquiry. Although he spent his life in Koningsberg, his ideas in epistemology, anthropology, ethics, and education gained global recognition and made him one of the most influential philosophers of the modern era. However, his views have also drawn criticism, especially for the inconsistencies that are perceived to support dogmatism, although he originally intended to criticize it. This research analyzes Kant's thought, especially about the limits of human knowledge, metaphysics, and criticism of dogmatism. The method used is qualitative text analysis, which examines Kant's main work, "Critique of Pure Reason," as well as the views of other philosophers. Primary and secondary literature is reviewed to identify inconsistencies in Kant's arguments. The results show that although Kant's philosophical system has made an outstanding contribution, there are significant areas for improvement in the handling of metaphysical concepts and the limits of knowledge, which ultimately has the potential to fall into dogmatism. This study concludes that Kant's thought remains relevant and essential, but it needs to be reread critically to fully understand its impact on modern philosophy.

On the existence of funneled orientations for classes of rooted phylogenetic networks

Recently, there has been a growing interest in the relationships between unrooted and rooted phylogenetic networks. In this context, a natural question to ask is if an unrooted phylogenetic network U can be oriented as a rooted phylogenetic network such that the latter satisfies certain structural properties. In a recent preprint, Bulteau et al. claim that it is NP-hard to decide if U has a funneled (resp. funneled tree-child) orientation, for when the internal vertices of U have degree at most 5. Unfortunately, the proof of their funneled tree-child result appears to be incorrect. In this paper, we show that, despite their incorrect proof, it is NP-hard to decide if U has a funneled tree-child orientation even if each internal vertex has degree 5 and that NP-hardness remains for other popular classes of rooted phylogenetic networks such as funneled normal and funneled reticulation-visible. Additionally, our results hold regardless of whether U is rooted at an existing vertex or by subdividing an edge with the root.

On the Mass of (Gravitational) Potential Energy

In classical mechanics, when a body freely moves or is externally forced to move in a conservative force field, such as a planet moving away from a star or a weight lifted from the floor, its kinetic energy or the work done on it is converted and stored as potential energy. The concept of potential energy was developed to uphold the fundamental principle of conservation of energy. According to the widely accepted interpretation of mass-energy equivalence, every form of energy has mass. This leads to the natural questions: does potential energy have mass? And if so, where is that mass located? We will start by briefly reviewing the issue through an examination of some key literature on the topic. The current consensus is that potential energy gets stored in the field energy of the interacting system. As a result of mass-energy equivalence, the equivalent mass is distributed throughout the entire space in some manner. However, this presents some difficulties. Here, like some other scholars in the past, we show that it contradicts the principles of special relativity and argue that potential energy does increase the mass of the bodies composing the system. We present an accessible thought experiment that heuristically corroborates that view specifically for the gravitational potential energy. We finally speculate on how that mass increase is distributed among the interacting bodies.