Search

The complexities of chats pose significant challenges for machine translation models. Recognizing the need for a precise evaluation metric to address the issues of chat translation, this study introduces Multidimensional Quality Metrics for Chat Translation (MQM-Chat). Through the experiments of five models using MQM-Chat, we observed that all models generated certain fundamental errors, while each of them has different shortcomings, such as omission, overly correcting ambiguous source content, and buzzword issues, resulting in the loss of stylized information. Our findings underscore the effectiveness of MQM-Chat in evaluating chat translation, emphasizing the importance of stylized content and dialogue consistency for future studies.

The complexities of chats pose significant challenges for machine translation models. Recognizing the need for a precise evaluation metric to address the issues of chat translation, this study introduces Multidimensional Quality Metrics for Chat Translation (MQM-Chat). Through the experiments of five models using MQM-Chat, we observed that all models generated certain fundamental errors, while each of them has different shortcomings, such as omission, overly correcting ambiguous source content, and buzzword issues, resulting in the loss of stylized information. Our findings underscore the effectiveness of MQM-Chat in evaluating chat translation, emphasizing the importance of stylized content and dialogue consistency for future studies.

Automated Short Answer Scoring (SAS) is the task of automatically scoring a given input to a prompt based on rubrics and reference answers. Although SAS is useful in real-world applications, both rubrics and reference answers differ between prompts, thus requiring a need to acquire new data and train a model for each new prompt. Such requirements are costly, especially for schools and online courses where resources are limited and only a few prompts are used. In this work, we attempt to reduce this cost through a two-phase approach: train a model on existing rubrics and answers with gold score signals and finetune it on a new prompt. Specifically, given that scoring rubrics and reference answers differ for each prompt, we utilize key phrases, or representative expressions that the answer should contain to increase scores, and train a SAS model to learn the relationship between key phrases and answers using already annotated prompts (i.e., cross-prompts). Our experimental results show that finetuning on existing cross-prompt data with key phrases significantly improves scoring accuracy, especially when the training data is limited. Finally, our extensive analysis shows that it is crucial to design the model so that it can learn the task's general property., Comment: This is the draft submitted to AIED 2023. For the latest version, please visit: https://link.springer.com/chapter/10.1007/978-3-031-36272-9_7

Gravitational waves (GWs) from gravitational three-body decay (graviton Bremsstrahlung process) can leave an indelible signal at ultrahigh frequencies. We focus on a scenario where superheavy particles are produced gravitationally at a transition between the inflationary and kination phases and analyze the detectability of the signal in the presence of GWs generated from the vacuum fluctuations during inflation. We find that, in many cases, GWs from the graviton Bremsstrahlung are buried in the stochastic gravitational wave background originating from inflation. However, if the superheavy particles are as heavy as the Planck scale, the graviton Bremsstrahlung can produce a sizable amount of GWs, surpassing the inflationary ones., Comment: 17 pages, 3 figures

Multi-step reasoning is widely adopted in the community to explore the better performance of language models (LMs). We report on the systematic strategy that LMs use in this process. Our controlled experiments reveal that LMs rely more heavily on heuristics, such as lexical overlap, in the earlier stages of reasoning when more steps are required to reach an answer. Conversely, as LMs progress closer to the final answer, their reliance on heuristics decreases. This suggests that LMs track only a limited number of future steps and dynamically combine heuristic strategies with logical ones in tasks involving multi-step reasoning.

Prior research in computational argumentation has mainly focused on scoring the quality of arguments, with less attention on explicating logical errors. In this work, we introduce four sets of explainable templates for common informal logical fallacies designed to explicate a fallacy's implicit logic. Using our templates, we conduct an annotation study on top of 400 fallacious arguments taken from LOGIC dataset and achieve a high agreement score (Krippendorf's alpha of 0.54) and reasonable coverage (0.83). Finally, we conduct an experiment for detecting the structure of fallacies and discover that state-of-the-art language models struggle with detecting fallacy templates (0.47 accuracy). To facilitate research on fallacies, we make our dataset and guidelines publicly available.

Language models (LMs) encode world knowledge in their internal parameters through training. However, LMs may learn personal and confidential information from the training data, leading to privacy concerns such as data leakage. Therefore, research on knowledge deletion from LMs is essential. This study focuses on the knowledge stored in LMs and analyzes the relationship between the side effects of knowledge deletion and the entities related to the knowledge. Our findings reveal that deleting knowledge related to popular entities can have catastrophic side effects. Furthermore, this research is the first to analyze knowledge deletion in models trained on synthetic knowledge graphs, indicating a new direction for controlled experiments.

Evaluating free-text explanations is a multifaceted, subjective, and labor-intensive task. Large language models (LLMs) present an appealing alternative due to their potential for consistency, scalability, and cost-efficiency. In this work, we present ACORN, a new dataset of 3,500 free-text explanations and aspect-wise quality ratings, and use it to gain insights into how LLMs evaluate explanations. We observed that replacing one of the human ratings sometimes maintained, but more often lowered the inter-annotator agreement across different settings and quality aspects, suggesting that their judgments are not always consistent with human raters. We further quantified this difference by comparing the correlation between LLM-generated ratings with majority-voted human ratings across different quality aspects. With the best system, Spearman's rank correlation ranged between 0.53 to 0.95, averaging 0.72 across aspects, indicating moderately high but imperfect alignment. Finally, we considered the alternative of using an LLM as an additional rater when human raters are scarce, and measured the correlation between majority-voted labels with a limited human pool and LLMs as an additional rater, compared to the original gold labels. While GPT-4 improved the outcome when there were only two human raters, in all other observed cases, LLMs were neutral to detrimental when there were three or more human raters. We publicly release the dataset to support future improvements in LLM-in-the-loop evaluation here: https://github.com/a-brassard/ACORN., Comment: 18 pages, 7 figures, under review. Data available here: https://github.com/a-brassard/ACORN

Coherent Ising Machine (CIM) is a network of optical parametric oscillators that solves combinatorial optimization problems by finding the ground state of an Ising Hamiltonian. As a practical application of CIM, Aonishi et al. proposed a quantum-classical hybrid system to solve optimization problems of L0-regularization-based compressed sensing (L0RBCS). Gunathilaka et al. has further enhanced the accuracy of the system. However, the computationally expensive CIM's stochastic differential equations (SDEs) limit the use of digital hardware implementations. As an alternative to Gunathilaka et al.'s CIM SDEs used previously, we propose using the mean-field CIM (MF-CIM) model, which is a physics-inspired heuristic solver without quantum noise. MF-CIM surmounts the high computational cost due to the simple nature of the differential equations (DEs). Furthermore, our results indicate that the proposed model has similar performance to physically accurate SDEs in both artificial and magnetic resonance imaging data, paving the way for implementing CIM-based L0RBCS on digital hardware such as Field Programmable Gate Arrays (FPGAs)., Comment: 19 pages, 7 figures

Supposing the Laser Interferometer Space Antenna (LISA) gravitational wave (GW) detector, we exhibit the detectability of a hypothetical smooth crossover in the early universe beyond the Standard Model of particle physics through the scalar-induced gravitational wave (SIGW) in terms of the Fisher forecast. A crossover at $\sim100\,\mathrm{TeV}$ can leave a signal on the GW spectrum in the $\sim\mathrm{mHz}$ frequency range, the sweet spot of the LISA sensitivity. These possibilities are also interesting in the primordial black hole (PBH) context as the associated PBH mass $\sim10^{22}\,\mathrm{g}$ lies at the window to explain the whole dark matter. We found that the properties of the crossover can be well determined if the power spectrum of primordial scalar perturbations are as large as $\sim5\times10^{-4}$ on the corresponding scale $\sim10^{12}\,\mathrm{Mpc^{-1}}$., Comment: 8 pages, 6 figures

Speakers sometimes omit certain arguments of a predicate in a sentence; such omission is especially frequent in pro-drop languages. This study addresses a question about ellipsis -- what can explain the native speakers' ellipsis decisions? -- motivated by the interest in human discourse processing and writing assistance for this choice. To this end, we first collect large-scale human annotations of whether and why a particular argument should be omitted across over 2,000 data points in the balanced corpus of Japanese, a prototypical pro-drop language. The data indicate that native speakers overall share common criteria for such judgments and further clarify their quantitative characteristics, e.g., the distribution of related linguistic factors in the balanced corpus. Furthermore, the performance of the language model-based argument ellipsis judgment model is examined, and the gap between the systems' prediction and human judgments in specific linguistic aspects is revealed. We hope our fundamental resource encourages further studies on natural human ellipsis judgment., Comment: 13 pages; accepted by LREC-COLING 2024

Mitigating the generation of contradictory responses poses a substantial challenge in dialogue response generation. The quality and quantity of available contradictory response data play a vital role in suppressing these contradictions, offering two significant benefits. First, having access to large contradiction data enables a comprehensive examination of their characteristics. Second, data-driven methods to mitigate contradictions may be enhanced with large-scale contradiction data for training. Nevertheless, no attempt has been made to build an extensive collection of model-generated contradictory responses. In this paper, we build a large dataset of response generation models' contradictions for the first time. Then, we acquire valuable insights into the characteristics of model-generated contradictions through an extensive analysis of the collected responses. Lastly, we also demonstrate how this dataset substantially enhances the performance of data-driven contradiction suppression methods., Comment: 16 pages

The magnetic properties of a circular graphene nanoribbon (carbon belt) in a magnetic field parallel to its central axis is studied using a tight-binding model. Orbital magnetic susceptibility is calculated using an analytical expression of the energy eigenvalues as a function of the magnetic flux density for any size, and its temperature dependence is considered. In the absence of electron hopping parallel to the magnetic field, the orbital magnetic susceptibility diverges at absolute zero if the chemical potential is zero and the number of atoms is a multiple of four. As the temperature increases, the magnitude of susceptibility decreases according to the power law, whose exponent depends on the size. In the presence of electron hopping parallel to the magnetic field, the divergence of the susceptibility near absolute zero disappears, and the sign changes with the transfer integral parallel to the magnetic field and the temperature., Comment: 20 pages, 7 figures

Language models (LMs) can express factual knowledge involving numeric properties such as Karl Popper was born in 1902. However, how this information is encoded in the model's internal representations is not understood well. Here, we introduce a simple method for finding and editing representations of numeric properties such as an entity's birth year. Empirically, we find low-dimensional subspaces that encode numeric properties monotonically, in an interpretable and editable fashion. When editing representations along directions in these subspaces, LM output changes accordingly. For example, by patching activations along a "birthyear" direction we can make the LM express an increasingly late birthyear: Karl Popper was born in 1929, Karl Popper was born in 1957, Karl Popper was born in 1968. Property-encoding directions exist across several numeric properties in all models under consideration, suggesting the possibility that monotonic representation of numeric properties consistently emerges during LM pretraining. Code: https://github.com/bheinzerling/numeric-property-repr

This paper proposes the task of automatic assessment of Sentence Translation Exercises (STEs), that have been used in the early stage of L2 language learning. We formalize the task as grading student responses for each rubric criterion pre-specified by the educators. We then create a dataset for STE between Japanese and English including 21 questions, along with a total of 3, 498 student responses (167 on average). The answer responses were collected from students and crowd workers. Using this dataset, we demonstrate the performance of baselines including finetuned BERT and GPT models with few-shot in-context learning. Experimental results show that the baseline model with finetuned BERT was able to classify correct responses with approximately 90% in F1, but only less than 80% for incorrect responses. Furthermore, the GPT models with few-shot learning show poorer results than finetuned BERT, indicating that our newly proposed task presents a challenging issue, even for the stateof-the-art large language models., Comment: 9 pages

A Gaussian quantum theory of bosonic modes has been widely used to describe quantum optical systems, including coherent Ising machines (CIMs) that consist of $\chi^{(2)}$ degenerate optical parametric oscillators (DOPOs) as nonlinear elements. However, Gaussian models have been thought to be invalid in the extremely strong-gain-saturation limit. Here, we develop an extended Gaussian model including two third-order fluctuation products, $\langle \delta \hat{X}^3\rangle$ and $\langle \delta \hat{X}\delta \hat{P}^2\rangle$, which we call self-skewness and cross-skewness, respectively. This new model which we call skew-Gaussian model more precisely replicates the success probability predicted by the quantum master equation (QME), relative to Gaussian models. We also discuss the impact of skew variables on the performance of CIMs.

Helices and spirals, prevalent across various systems, play a crucial role in characterizing symmetry, describing dynamics, and imparting unique functionalities, attributed to their inherent simplicity and chiral nature. A helical excitation on a quantized vortex, an example of a one-dimensional topological defect, emerges as a Nambu-Goldstone mode following spontaneous symmetry breaking, known as a Kelvin wave. Kelvin waves play a vital role in energy dissipation within inviscid quantum fluids. However, deliberately exciting Kelvin waves has proven to be challenging. Here, we introduce a controlled method for exciting Kelvin waves on a quantized vortex in superfluid helium-4. We used a charged nanoparticle, oscillated by a time-varying electric field, to stimulate Kelvin waves on the vortex. A major breakthrough in our research is the confirmation of the helical nature of Kelvin waves through three-dimensional image reconstruction, providing visual evidence of their complex dynamics. Additionally, we determined the dispersion relation and the phase velocity of the Kelvin wave and identified the vorticity direction, enhancing our understanding of quantum fluid behavior. This work elucidates the dynamics of Kelvin waves and pioneers a novel approach for manipulating and observing quantized vortices in three dimensions, thereby opening new avenues for exploring quantum fluidic systems., Comment: 18 pages, 4 figures

Solving inverse problems to identify Hamiltonians with desired properties holds promise for the discovery of fundamental principles. In quantum systems, quantum entanglement plays a pivotal role in not only characterizing the quantum nature but also developing quantum technology like quantum computing. Nonetheless, the design principles of the quantum entanglement are yet to be clarified. Here we apply an inverse design framework using automatic differentiation to quantum spin systems, aiming to construct Hamiltonians with large quantum entanglement. We show that the method automatically finds the Kitaev model with bond-dependent anisotropic interactions, whose ground state is a quantum spin liquid, on both honeycomb and square-octagon lattices. On triangular and maple-leaf lattices with geometrical frustration, it generates numerous solutions with spatially inhomogeneous interactions rather than converging to a specific model, but it still helps to construct unprecedented models. The comparative study reveals that bond-dependent anisotropic interactions, rather than isotropic Heisenberg interactions, amplify quantum entanglement, even in systems with geometrical frustration. The present study paves the way for the automatic design of new quantum systems with desired quantum nature and functionality.

We introduce a Japanese Morphology dataset, J-UniMorph, developed based on the UniMorph feature schema. This dataset addresses the unique and rich verb forms characteristic of the language's agglutinative nature. J-UniMorph distinguishes itself from the existing Japanese subset of UniMorph, which is automatically extracted from Wiktionary. On average, the Wiktionary Edition features around 12 inflected forms for each word and is primarily dominated by denominal verbs (i.e., [noun] +suru (do-PRS)). Morphologically, this form is equivalent to the verb suru (do). In contrast, J-UniMorph explores a much broader and more frequently used range of verb forms, offering 118 inflected forms for each word on average. It includes honorifics, a range of politeness levels, and other linguistic nuances, emphasizing the distinctive characteristics of the Japanese language. This paper presents detailed statistics and characteristics of J-UniMorph, comparing it with the Wiktionary Edition. We release J-UniMorph and its interactive visualizer publicly available, aiming to support cross-linguistic research and various applications., Comment: 14 pages, 4 figures

The subsolar mass primordial black hole (PBH) attracts attention as robust evidence of its primordial origin against the astrophysical black hole. Not only with themselves, PBHs can also form binaries with ordinary astrophysical objects, catching them by gravitational wave (GW) bremsstrahlung. We discuss the detectability of the inspiral GWs from binaries consisting of a PBH and a white dwarf (WD) by using space-borne gravitational wave interferometers like DECIGO. The conservative assessment shows the expected event number in three years by DECIGO is $\mathcal{O}(10^{-6})$ for $M_\mathrm{PBH} \sim 0.1M_\odot$. Possible enhancement mechanisms of WD-PBH binary formation may amplify this event rate. We discuss how large enhancement associated with WDs is required to detect WD-PBH merger events without violating the existing constraints on the PBH-PBH merger by the ground-based detector., Comment: 13 pages, 10 figures

The scalar-induced gravitational waves (SIGW), arising from large amplitude primordial density fluctuations, provide a unique observational test for directly probing the epoch of inflation. In this work, we provide constraints on the SIGW background by taking into account the non-Gaussianity in the primordial density fluctuations, using the third observing run (O3) data of the LIGO-Virgo-KAGRA collaboration. We find that the non-Gaussianity gives a non-negligible effect on the GW energy density spectrum and starts to affect the analysis of the O3 data when the non-Gaussianity parameter is $F_{\rm NL} > 3.55$. Furthermore, the constraints exhibit asymptotic behavior given by $F_{\rm NL} A_g = \rm{const.}$ at large $F_{\rm NL}$ limit, where $A_g$ denotes the amplitude of the curvature perturbations. In the limit of large $F_{\rm NL}$, we placed a 95% confidence level upper limit $F_{\rm NL} A_g \leq 0.13, 0.09, 0.10$ at fixed scales of $10^{16}, 10^{16.5}, 10^{17}~{\rm Mpc}^{-1}$, respectively., Comment: 17 pages, 4 figures. Changes match published version

Squared deviations from the equilibrium positions of one-dimensional coupled harmonic oscillators with fixed and free endpoints are calculated, and the time averages are expressed as a function of the initial displacements and velocities. Furthermore, we consider the averages of squared deviations over an ensemble of initial displacements and velocities, which distribute based on a product of the same distribution functions with variances of the initial coordinates $\sigma_{x}^2$ and velocities $\sigma_{v}^2$, respectively. We demonstrate that the mean squared deviation linearly increases as the oscillator separates further from the fixed endpoint because of the asymmetrical boundary conditions, and that the increase rate depends only on $\sigma_{v}^2$ and not on $\sigma_{x}^2$. This simple statistical property of harmonic oscillation is similarly observed in the oscillations of a graphene sheet and carbon nanotube in molecular dynamics simulations, in which the interacting forces are nonlinear., Comment: 21 pages, 10 figures

Factual probing is a method that uses prompts to test if a language model "knows" certain world knowledge facts. A problem in factual probing is that small changes to the prompt can lead to large changes in model output. Previous work aimed to alleviate this problem by optimizing prompts via text mining or fine-tuning. However, such approaches are relation-specific and do not generalize to unseen relation types. Here, we propose to use test-time augmentation (TTA) as a relation-agnostic method for reducing sensitivity to prompt variations by automatically augmenting and ensembling prompts at test time. Experiments show improved model calibration, i.e., with TTA, model confidence better reflects prediction accuracy. Improvements in prediction accuracy are observed for some models, but for other models, TTA leads to degradation. Error analysis identifies the difficulty of producing high-quality prompt variations as the main challenge for TTA., Comment: 12 pages, 4 figures, accepted to EMNLP 2023 Findings (short paper)

The performance of sentence encoders can be significantly improved through the simple practice of fine-tuning using contrastive loss. A natural question arises: what characteristics do models acquire during contrastive learning? This paper theoretically and experimentally shows that contrastive-based sentence encoders implicitly weight words based on information-theoretic quantities; that is, more informative words receive greater weight, while others receive less. The theory states that, in the lower bound of the optimal value of the contrastive learning objective, the norm of word embedding reflects the information gain associated with the distribution of surrounding words. We also conduct comprehensive experiments using various models, multiple datasets, two methods to measure the implicit weighting of models (Integrated Gradients and SHAP), and two information-theoretic quantities (information gain and self-information). The results provide empirical evidence that contrastive fine-tuning emphasizes informative words., Comment: 16 pages, 6 figures, accepted to EMNLP 2023 Findings (short paper)

We consider the one-dimensional nonlinear Schr\"odinger equation with focusing, power nonlinearity, and a repulsive delta potential. We show that if the potential is not too strong, the construction by Nguy\~{\^e}n (2019) of solutions converging strongly at time infinity to a pair of logarithmically separating solitons can be adapted to accommodate the effect of the potential. On the other hand, we show that if the potential is stronger, no such solutions exist.

We consider the $L^2$-supercritical nonlinear Schr\"{o}dinger equation with a repulsive Dirac delta potential in one dimensional space. In a previous work, we clarified the global dynamics of even solutions with the same action as the high-frequency ground state standing wave solutions. In that case, there are obvious non-scattering global solutions, i.e., the standing waves. In this paper, we show a scattering and blow-up dichotomy for threshold even solutions in the low-frequency case. We emphasize that this dichotomy still holds at the critical frequency between high and low.

We prove the existence of multi-soliton solutions for the nonlinear Schr\"{o}dinger equation with repulsive Dirac delta potential and $L^2$-supercritical focusing nonlinear term. Our main contribution is to treat the unmoving part of the multi-solitons, which is the ground state of the equation. The linearized operator around it has two unstable eigenvalues. This is the main difference from NLS without potential, whose existence of multi-solitons is investigated by C\^{o}te, Martel, and Merle (2011).

Superfluid helium-4 (He II) is a widely adopted coolant in scientific and engineering applications owing to its exceptional heat transfer capabilities. However, boiling can spontaneously occur on a heating surface in He II when the heat flux exceeds a threshold value $q^*$, referred to as the peak heat flux. While the parameter $q^*$ holds paramount importance in the design of He II based cooling systems, extensive research has primarily focused on its behavior in steady homogeneous heat transfer from a flat heating surface. For inhomogeneous heat transfer from curved surfaces, $q^*$ exhibits intricate dependance on parameters such as the He II bath temperature $T_b$, the immersion depth $h$, and the curvature radius $R_0$ of the heating surface. A comprehensive understanding on how $q^*$ depends on these parameters remains elusive. In this paper, we report our systematic study on $q^*$ for steady heat transfer from cylindrical and spherical heaters in He II. We compute $q^*$ for a wide range of parameter combinations $(T_b, h, R_0)$ by solving the He II two-fluid equations of motion. The generated data have allowed us to develop a robust correlation that accurately reproduces $q^*$ for all the parameter combinations we explored. Our findings, particularly the establishment of the correlation, carry valuable implications for emergent applications that involve steady inhomogeneous heat transfer in He II systems., Comment: 10 pages, 10 figures

Coherent Ising Machine (CIM) is a network of optical parametric oscillators that solves combinatorial optimization problems by finding the ground state of an Ising Hamiltonian. In CIMs, a problem arises when attempting to realize the Zeeman term because of the mismatch in size between interaction and Zeeman terms due to the variable amplitude of the optical parametric oscillator pulses corresponding to spins. There have been three approaches proposed so far to address this problem for CIM, including the absolute mean amplitude method, the auxiliary spin method, and the chaotic amplitude control (CAC) method. This paper focuses on the efficient implementation of Zeeman terms within the mean-field CIM model, which is a physics-inspired heuristic solver without quantum noise. With the mean-field model, computation is easier than with more physically accurate models, which makes it suitable for implementation in field programmable gate arrays (FPGAs) and large-scale simulations. Firstly, we examined the performance of the mean-field CIM model for realizing the Zeeman term with the CAC method, as well as their performance when compared to a more physically accurate model. Next, we compared the CAC method to other Zeeman term realization techniques on the mean-field model and a more physically accurate model. In both models, the CAC method outperformed the other methods while retaining similar performance., Comment: 8 pages, 4 figures

We show that the recently reported cosmic gravitational wave background by the NANOGrav 15-year collaboration may be the result of resonant particle creation during inflation. For the appropriate amplitude and particle mass an enhancement of the primordial scalar power spectrum could induce Secondary Induced Gravitational Waves (SIGW) which will appear on a scale corresponding to the frequency of the NANOGrav detection. Since the resonant creation will have an effect comparable to that of a delta function increment as studied by the NANOGrav 15-year collaboration, our study indicates that the low-frequency Pulsar Timing Array (PTA) data could reveal the aspects of the physics during inflation through the detection of a cosmic background of Gravitational Waves (GW)., Comment: 10 pages, 8 figures

Nonlocal spin polarization phenomena are thoroughly investigated in the devices made of chiral metallic single crystals of CrNb$_3$S$_6$ and NbSi$_2$ as well as of polycrystalline NbSi$_2$. We demonstrate that simultaneous injection of charge currents in the opposite ends of the device with the nonlocal setup induces the switching behavior of spin polarization in a controllable manner. Such a nonlocal spin polarization appears regardless of the difference in the materials and device dimensions, implying that the current injection in the nonlocal configuration splits spin-dependent chemical potentials throughout the chiral crystal even though the current is injected into only a part of the crystal. We show that the proposed model of the spin dependent chemical potentials explains the experimental data successfully. The nonlocal double-injection device may offer significant potential to control the spin polarization to large areas because of the nature of long-range nonlocal spin polarization in chiral materials., Comment: 8 pages, 8 figures

In this paper, we describe the development of a communication support system that detects erroneous translations to facilitate crosslingual communications due to the limitations of current machine chat translation methods. We trained an error detector as the baseline of the system and constructed a new Japanese-English bilingual chat corpus, BPersona-chat, which comprises multiturn colloquial chats augmented with crowdsourced quality ratings. The error detector can serve as an encouraging foundation for more advanced erroneous translation detection systems.

Quantum computers (QCs), which work based on the law of quantum mechanics, are expected to be faster than classical computers in several computational tasks such as prime factoring and simulation of quantum many-body systems. In the last decade, research and development of QCs have rapidly advanced. Now hundreds of physical qubits are at our disposal, and one can find several remarkable experiments actually outperforming the classical computer in a specific computational task. On the other hand, it is unclear what the typical usages of the QCs are. Here we conduct an extensive survey on the papers that are posted in the quant-ph section in arXiv and claim to have used QCs in their abstracts. To understand the current situation of the research and development of the QCs, we evaluated the descriptive statistics about the papers, including the number of qubits employed, QPU vendors, application domains and so on. Our survey shows that the annual number of publications is increasing, and the typical number of qubits employed is about six to ten, growing along with the increase in the quantum volume (QV). Most of the preprints are devoted to applications such as quantum machine learning, condensed matter physics, and quantum chemistry, while quantum error correction and quantum noise mitigation use more qubits than the other topics. These imply that the increase in QV is fundamentally relevant, and more experiments for quantum error correction, and noise mitigation using shallow circuits with more qubits will take place., Comment: 14 pages, 5 figures, figures regenerated

The use of argumentation in education has been shown to improve critical thinking skills for end-users such as students, and computational models for argumentation have been developed to assist in this process. Although these models are useful for evaluating the quality of an argument, they oftentimes cannot explain why a particular argument is considered poor or not, which makes it difficult to provide constructive feedback to users to strengthen their critical thinking skills. In this survey, we aim to explore the different dimensions of feedback (Richness, Visualization, Interactivity, and Personalization) provided by the current computational models for argumentation, and the possibility of enhancing the power of explanations of such models, ultimately helping learners improve their critical thinking skills., Comment: 14 pages, 4 figures

Prediction head is a crucial component of Transformer language models. Despite its direct impact on prediction, this component has often been overlooked in analyzing Transformers. In this study, we investigate the inner workings of the prediction head, specifically focusing on bias parameters. Our experiments with BERT and GPT-2 models reveal that the biases in their word prediction heads play a significant role in the models' ability to reflect word frequency in a corpus, aligning with the logit adjustment method commonly used in long-tailed learning. We also quantify the effect of controlling the biases in practical auto-regressive text generation scenarios; under a particular setting, more diverse text can be generated without compromising text quality., Comment: 11 pages, 12 figures, accepted to ACL 2023 Findings (short paper)