Hitomi Yanaka

Jin Tanaka, Daiki Matsuoka, Ryoma Kumon et al.

We investigate the extent to which the language processing of LLMs resembles human cognitive processes, focusing on a human cognitive bias called the $\textit{neglect-zero effect}$. This effect refers to the human tendency to ignore $\textit{zero-models}$, which are configurations that render a proposition vacuously true by virtue of an empty set. We focus on two types of inferences driven by the neglect-zero effect, and examine how LLMs process these inferences by comparing their behavior with that in an inference that does not involve the neglect-zero effect. For this purpose, we employ a paradigm based on $\textit{structural priming}$, where recent exposure to a preceding sentence (the $\textit{prime}$) facilitates the processing of a subsequent sentence (the $\textit{target}$) due to their structural similarity. We prepare primes to force LLMs to consider the zero-model, and analyze whether they also consider it in the target. The results suggest that the neglect-zero effect may not occur in the LLMs analyzed in this study. Our code is available at https://github.com/ynklab/neglect_zero

LLM-jp, Akiko Aizawa, Eiji Aramaki et al.

This paper introduces LLM-jp, a cross-organizational project for the research and development of Japanese large language models (LLMs). LLM-jp aims to develop open-source and strong Japanese LLMs, and as of this writing, more than 1,500 participants from academia and industry are working together for this purpose. This paper presents the background of the establishment of LLM-jp, summaries of its activities, and technical reports on the LLMs developed by LLM-jp. For the latest activities, visit https://llm-jp.nii.ac.jp/en/.

Wataru Nakata, Kentaro Seki, Hitomi Yanaka et al.

Spoken dialogue plays a crucial role in human-AI interactions, necessitating dialogue-oriented spoken language models (SLMs). To develop versatile SLMs, large-scale and diverse speech datasets are essential. Additionally, to ensure hiqh-quality speech generation, the data must be spontaneous like in-wild data and must be acoustically clean with noise removed. Despite the critical need, no open-source corpus meeting all these criteria has been available. This study addresses this gap by constructing and releasing a large-scale spoken dialogue corpus, named Japanese Corpus for Human-AI Talks (J-CHAT), which is publicly accessible. Furthermore, this paper presents a language-independent method for corpus construction and describes experiments on dialogue generation using SLMs trained on J-CHAT. Experimental results indicate that the collected data from multiple domains by our method improve the naturalness and meaningfulness of dialogue generation.

Koki Ryu, Hitomi Yanaka

Personalized image aesthetics assessment (PIAA) is an important research problem with practical real-world applications. While methods based on vision-language models (VLMs) are promising candidates for PIAA, it remains unclear whether they internally encode rich, multi-level aesthetic attributes required for effective personalization. In this paper, we first analyze the internal representations of VLMs to examine the presence and distribution of such aesthetic attributes, and then leverage them for lightweight, individual-level personalization without model fine-tuning. Our analysis reveals that VLMs encode diverse aesthetic attributes that propagate into the language decoder layers. Building on these representations, we demonstrate that simple linear models can perform PIAA effectively. We further analyze how aesthetic information is transferred across layers in different VLM architectures and across image domains. Our findings provide insights into how VLMs can be utilized for modeling subjective, individual aesthetic preferences. Our code is available at https://github.com/ynklab/vlm-latent-piaa.

Ryo Sekizawa, Nan Duan, Shuai Lu et al.

Code search is a task to find programming codes that semantically match the given natural language queries. Even though some of the existing datasets for this task are multilingual on the programming language side, their query data are only in English. In this research, we create a multilingual code search dataset in four natural and four programming languages using a neural machine translation model. Using our dataset, we pre-train and fine-tune the Transformer-based models and then evaluate them on multiple code search test sets. Our results show that the model pre-trained with all natural and programming language data has performed best in most cases. By applying back-translation data filtering to our dataset, we demonstrate that the translation quality affects the model's performance to a certain extent, but the data size matters more.

Hitomi Yanaka, Koji Mineshima

Natural Language Inference (NLI) and Semantic Textual Similarity (STS) are widely used benchmark tasks for compositional evaluation of pre-trained language models. Despite growing interest in linguistic universals, most NLI/STS studies have focused almost exclusively on English. In particular, there are no available multilingual NLI/STS datasets in Japanese, which is typologically different from English and can shed light on the currently controversial behavior of language models in matters such as sensitivity to word order and case particles. Against this background, we introduce JSICK, a Japanese NLI/STS dataset that was manually translated from the English dataset SICK. We also present a stress-test dataset for compositional inference, created by transforming syntactic structures of sentences in JSICK to investigate whether language models are sensitive to word order and case particles. We conduct baseline experiments on different pre-trained language models and compare the performance of multilingual models when applied to Japanese and other languages. The results of the stress-test experiments suggest that the current pre-trained language models are insensitive to word order and case marking.

Adam Nohejl, Xuanxin Wu, Yusuke Ide et al.

We describe two types of models for vocabulary difficulty prediction: a high-accuracy black-box model, which achieved the top shared task result in the open track, and an explainable model, which outperforms a fine-tuned encoder baseline. As the black-box model, we fine-tuned an LLM using a soft-target loss function for effective application to the rating task, achieving r > 0.91. The explainable model provides insights into what impacts the difficulty of each item while maintaining a strong correlation (r > 0.77). We further analyze the results, demonstrating that the difficulty of items in the British Council's Knowledge-based Vocabulary Lists (KVL) is often affected by spelling difficulty or the construction of the test items, in addition to the genuine production difficulty of the words. We make our code available online at https://github.com/adno/vocabulary-difficulty .

Tomoki Sugimoto, Yasumasa Onoe, Hitomi Yanaka

Natural Language Inference (NLI) tasks involving temporal inference remain challenging for pre-trained language models (LMs). Although various datasets have been created for this task, they primarily focus on English and do not address the need for resources in other languages. It is unclear whether current LMs realize the generalization capacity for temporal inference across languages. In this paper, we present Jamp, a Japanese NLI benchmark focused on temporal inference. Our dataset includes a range of temporal inference patterns, which enables us to conduct fine-grained analysis. To begin the data annotation process, we create diverse inference templates based on the formal semantics test suites. We then automatically generate diverse NLI examples by using the Japanese case frame dictionary and well-designed templates while controlling the distribution of inference patterns and gold labels. We evaluate the generalization capacities of monolingual/multilingual LMs by splitting our dataset based on tense fragments (i.e., temporal inference patterns). Our findings demonstrate that LMs struggle with specific linguistic phenomena, such as habituality, indicating that there is potential for the development of more effective NLI models across languages.

Ryoma Kumon, Hitomi Yanaka

While language models demonstrate sophisticated syntactic capabilities, the extent to which their internal mechanisms align with cross-constructional principles studied in linguistics remains poorly understood. This study investigates whether models employ shared neural mechanisms across different syntactic constructions by applying causal interpretability methods at a granular level. Focusing on filler-gap dependencies and negative polarity item (NPI) licensing, we utilize activation patching to identify the functional roles of specific attention heads and MLP blocks. Our results reveal a highly localized and shared mechanism for filler-gap dependencies located in the early to middle layers, whereas NPI processing exhibits no such unified mechanism. Furthermore, we find that these mechanisms identified by activation patching generalize to out-of-distribution, while distributed alignment search, a supervised interpretability method, is susceptible to overfitting on narrow linguistic distributions. Finally, we validate our findings by demonstrating that the manipulation of the identified components improves model performance on acceptability judgment benchmarks.

Tomoki Sugimoto, Hitomi Yanaka

Natural Language Inference (NLI) is the task of determining whether a premise entails a hypothesis. NLI with temporal order is a challenging task because tense and aspect are complex linguistic phenomena involving interactions with temporal adverbs and temporal connectives. To tackle this, temporal and aspectual inference has been analyzed in various ways in the field of formal semantics. However, a Japanese NLI system for temporal order based on the analysis of formal semantics has not been sufficiently developed. We present a logic-based NLI system that considers temporal order in Japanese based on compositional semantics via Combinatory Categorial Grammar (CCG) syntactic analysis. Our system performs inference involving temporal order by using axioms for temporal relations and automated theorem provers. We evaluate our system by experimenting with Japanese NLI datasets that involve temporal order. We show that our system outperforms previous logic-based systems as well as current deep learning-based models.

Qingcheng Zeng, Yuheng Lu, Zeqi Zhou et al.

Code-switching is a pervasive linguistic phenomenon in global communication, yet modern information retrieval systems remain predominantly designed for, and evaluated within, monolingual contexts. To bridge this critical disconnect, we present a holistic study dedicated to code-switching IR. We introduce CSR-L (Code-Switching Retrieval benchmark-Lite), constructing a dataset via human annotation to capture the authentic naturalness of mixed-language queries. Our evaluation across statistical, dense, and late-interaction paradigms reveals that code-switching acts as a fundamental performance bottleneck, degrading the effectiveness of even robust multilingual models. We demonstrate that this failure stems from substantial divergence in the embedding space between pure and code-switched text. Scaling this investigation, we propose CS-MTEB, a comprehensive benchmark covering 11 diverse tasks, where we observe performance declines of up to 27%. Finally, we show that standard multilingual techniques like vocabulary expansion are insufficient to resolve these deficits completely. These findings underscore the fragility of current systems and establish code-switching as a crucial frontier for future IR optimization.

Daisuke Bekki, Hitomi Yanaka

The Japanese CCGBank serves as training and evaluation data for developing Japanese CCG parsers. However, since it is automatically generated from the Kyoto Corpus, a dependency treebank, its linguistic validity still needs to be sufficiently verified. In this paper, we focus on the analysis of passive/causative constructions in the Japanese CCGBank and show that, together with the compositional semantics of ccg2lambda, a semantic parsing system, it yields empirically wrong predictions for the nested construction of passives and causatives.

Tomoya Kurosawa, Hitomi Yanaka

Recently, the Natural Language Inference (NLI) task has been studied for semi-structured tables that do not have a strict format. Although neural approaches have achieved high performance in various types of NLI, including NLI between semi-structured tables and texts, they still have difficulty in performing a numerical type of inference, such as counting. To handle a numerical type of inference, we propose a logical inference system for reasoning between semi-structured tables and texts. We use logical representations as meaning representations for tables and texts and use model checking to handle a numerical type of inference between texts and tables. To evaluate the extent to which our system can perform inference with numerical comparatives, we make an evaluation protocol that focuses on numerical understanding between semi-structured tables and texts in English. We show that our system can more robustly perform inference between tables and texts that requires numerical understanding compared with current neural approaches.

Amane Watahiki, Tomoki Doi, Akari Kikuchi et al.

Narrative analysis is a cornerstone of qualitative research. One leading approach is the Labovian model, but its application is labor-intensive, requiring a holistic, recursive interpretive process that moves back and forth between individual parts of the transcript and the transcript as a whole. Existing Labovian datasets are available only in English, which differs markedly from Japanese in terms of grammar and discourse conventions. To address this gap, we introduce the first systematic guidelines for Labovian narrative analysis of Japanese narrative data. Our guidelines retain all six Labovian categories and extend the framework by providing explicit rules for clause segmentation tailored to Japanese constructions. In addition, our guidelines cover a broader range of clause types and narrative types. Using these guidelines, annotators achieved high agreement in clause segmentation (Fleiss' kappa = 0.80) and moderate agreement in two structural classification tasks (Krippendorff's alpha = 0.41 and 0.45, respectively), one of which is slightly higher than that found in prior work despite the use of finer-grained distinctions. This paper describes the Labovian model, the proposed guidelines, the annotation process, and their utility. It concludes by discussing the challenges encountered during the annotation process and the prospects for developing a larger dataset for structural narrative analysis in Japanese qualitative research.

Yusuke Ide, Adam Nohejl, Joshua Tanner et al.

We study dictionary definition generation (DDG), i.e., the generation of non-contextualized definitions for given headwords. Dictionary definitions are an essential resource for learning word senses, but manually creating them is costly, which motivates us to automate the process. Specifically, we address learner's dictionary definition generation (LDDG), where definitions should consist of simple words. First, we introduce a reliable evaluation approach for DDG, based on our new evaluation criteria and powered by an LLM-as-a-judge. To provide reference definitions for the evaluation, we also construct a Japanese dataset in collaboration with a professional lexicographer. Validation results demonstrate that our evaluation approach agrees reasonably well with human annotators. Second, we propose an LDDG approach via iterative simplification with an LLM. Experimental results indicate that definitions generated by our approach achieve high scores on our criteria while maintaining lexical simplicity.

Tomoki Doi, Masaru Isonuma, Hitomi Yanaka

Large language models have the potential to generate explanations for their own predictions in a variety of styles based on user instructions. Recent research has examined whether these self-explanations faithfully reflect the models' actual behavior and has found that they often lack faithfulness. However, the question of how to improve faithfulness remains underexplored. Moreover, because different explanation styles have superficially distinct characteristics, it is unclear whether improvements observed in one style also arise when using other styles. This study analyzes the effects of training for faithful self-explanations and the extent to which these effects generalize, using three classification tasks and three explanation styles. We construct one-word constrained explanations that are likely to be faithful using a feature attribution method, and use these pseudo-faithful self-explanations for continual learning on instruction-tuned models. Our experiments demonstrate that training can improve self-explanation faithfulness across all classification tasks and explanation styles, and that these improvements also show signs of generalization to the multi-word settings and to unseen tasks. Furthermore, we find consistent cross-style generalization among three styles, suggesting that training may contribute to a broader improvement in faithful self-explanation ability.

Hirohane Takagi, Gouki Minegishi, Shota Kizawa et al.

Although behavioral studies have documented numerical reasoning errors in large language models (LLMs), the underlying representational mechanisms remain unclear. We hypothesize that numerical attributes occupy shared latent subspaces and investigate two questions:(1) How do LLMs internally integrate multiple numerical attributes of a single entity? (2)How does irrelevant numerical context perturb these representations and their downstream outputs? To address these questions, we combine linear probing with partial correlation analysis and prompt-based vulnerability tests across models of varying sizes. Our results show that LLMs encode real-world numerical correlations but tend to systematically amplify them. Moreover, irrelevant context induces consistent shifts in magnitude representations, with downstream effects that vary by model size. These findings reveal a vulnerability in LLM decision-making and lay the groundwork for fairer, representation-aware control under multi-attribute entanglement.

Ryo Sekizawa, Hitomi Yanaka

Using Japanese honorifics is challenging because it requires not only knowledge of the grammatical rules but also contextual information, such as social relationships. It remains unclear whether pre-trained large language models (LLMs) can flexibly handle Japanese honorifics like humans. To analyze this, we introduce an honorific conversion task that considers social relationships among people mentioned in a conversation. We construct a Japanese honorifics dataset from problem templates of various sentence structures to investigate the syntactic generalization capacity of GPT-3, one of the leading LLMs, on this task under two settings: fine-tuning and prompt learning. Our results showed that the fine-tuned GPT-3 performed better in a context-aware honorific conversion task than the prompt-based one. The fine-tuned model demonstrated overall syntactic generalizability towards compound honorific sentences, except when tested with the data involving direct speech.

Tomoya Kurosawa, Hitomi Yanaka

Even in the era of massive language models, it has been suggested that character-level representations improve the performance of neural models. The state-of-the-art neural semantic parser for Discourse Representation Structures uses character-level representations, improving performance in the four languages (i.e., English, German, Dutch, and Italian) in the Parallel Meaning Bank dataset. However, how and why character-level information improves the parser's performance remains unclear. This study provides an in-depth analysis of performance changes by order of character sequences. In the experiments, we compare F1-scores by shuffling the order and randomizing character sequences after testing the performance of character-level information. Our results indicate that incorporating character-level information does not improve the performance in English and German. In addition, we find that the parser is not sensitive to correct character order in Dutch. Nevertheless, performance improvements are observed when using character-level information.

Taisei Yamamoto, Ryoma Kumon, Danushka Bollegala et al.

As large language models (LLMs) are increasingly deployed worldwide, ensuring their fair and comprehensive cultural understanding is important. However, LLMs exhibit cultural bias and limited awareness of underrepresented cultures, while the mechanisms underlying their cultural understanding remain underexplored. To fill this gap, we conduct a neuron-level analysis to identify neurons that drive cultural behavior, introducing a gradient-based scoring method with additional filtering for precise refinement. We identify both culture-general neurons contributing to cultural understanding regardless of cultures, and culture-specific neurons tied to an individual culture. These neurons account for less than 1% of all neurons and are concentrated in shallow to middle MLP layers. We validate their role by showing that suppressing them substantially degrades performance on cultural benchmarks (by up to 30%), while performance on general natural language understanding (NLU) benchmarks remains largely unaffected. Moreover, we show that culture-specific neurons support knowledge of not only the target culture, but also related cultures. Finally, we demonstrate that training on NLU benchmarks can diminish models' cultural understanding when we update modules containing many culture-general neurons. These findings provide insights into the internal mechanisms of LLMs and offer practical guidance for model training and engineering. Our code is available at https://github.com/ynklab/CULNIG

Koki Ryu, Hitomi Yanaka

Personalizing the outputs of large language models (LLMs) to align with individual user preferences is an active research area. However, previous studies have mainly focused on classification or ranking tasks and have not considered Likert-scale rating prediction, a regression task that requires both language and mathematical reasoning to be solved effectively. This task has significant industrial applications, but the utilization of LLMs remains underexplored, particularly regarding the capabilities of off-the-shelf LLMs. This study investigates the performance of off-the-shelf LLMs on rating prediction, providing different in-context information. Through comprehensive experiments with eight models across three datasets, we demonstrate that user-written reviews significantly improve the rating prediction performance of LLMs. This result is comparable to traditional methods like matrix factorization, highlighting the potential of LLMs as a promising solution for the cold-start problem. We also find that the reviews for concrete items are more effective than general preference descriptions that are not based on any specific item. Furthermore, we discover that prompting LLMs to first generate a hypothetical review enhances the rating prediction performance. Our code is available at https://github.com/ynklab/rating-prediction-with-reviews.

Jie Lu, Du Jin, Hitomi Yanaka

Unlike English, which uses distinct forms (e.g., had, has, will have) to mark the perfect aspect across tenses, Chinese and Japanese lack separate grammatical forms for tense within the perfect aspect, which complicates Natural Language Inference (NLI). Focusing on the perfect aspect in these languages, we construct a linguistically motivated, template-based NLI dataset (1,350 pairs per language). Experiments reveal that even advanced LLMs struggle with temporal inference, particularly in detecting subtle tense and reference-time shifts. These findings highlight model limitations and underscore the need for cross-linguistic evaluation in temporal semantics. Our dataset is available at https://github.com/Lujie2001/CrossNLI.

Hitomi Yanaka, Namgi Han, Ryoma Kumon et al.

With the development of large language models (LLMs), social biases in these LLMs have become a pressing issue. Although there are various benchmarks for social biases across languages, the extent to which Japanese LLMs exhibit social biases has not been fully investigated. In this study, we construct the Japanese Bias Benchmark dataset for Question Answering (JBBQ) based on the English bias benchmark BBQ, with analysis of social biases in Japanese LLMs. The results show that while current open Japanese LLMs with more parameters show improved accuracies on JBBQ, their bias scores increase. In addition, prompts with a warning about social biases and chain-of-thought prompting reduce the effect of biases in model outputs, but there is room for improvement in extracting the correct evidence from contexts in Japanese. Our dataset is available at https://github.com/ynklab/JBBQ_data.

Hitomi Yanaka, Koji Mineshima, Kentaro Inui

Recently, deep neural networks (DNNs) have achieved great success in semantically challenging NLP tasks, yet it remains unclear whether DNN models can capture compositional meanings, those aspects of meaning that have been long studied in formal semantics. To investigate this issue, we propose a Systematic Generalization testbed based on Natural language Semantics (SyGNS), whose challenge is to map natural language sentences to multiple forms of scoped meaning representations, designed to account for various semantic phenomena. Using SyGNS, we test whether neural networks can systematically parse sentences involving novel combinations of logical expressions such as quantifiers and negation. Experiments show that Transformer and GRU models can generalize to unseen combinations of quantifiers, negations, and modifiers that are similar to given training instances in form, but not to the others. We also find that the generalization performance to unseen combinations is better when the form of meaning representations is simpler. The data and code for SyGNS are publicly available at https://github.com/verypluming/SyGNS.

Hitomi Yanaka, Koji Mineshima, Kentaro Inui

Despite the recent success of deep neural networks in natural language processing, the extent to which they can demonstrate human-like generalization capacities for natural language understanding remains unclear. We explore this issue in the domain of natural language inference (NLI), focusing on the transitivity of inference relations, a fundamental property for systematically drawing inferences. A model capturing transitivity can compose basic inference patterns and draw new inferences. We introduce an analysis method using synthetic and naturalistic NLI datasets involving clause-embedding verbs to evaluate whether models can perform transitivity inferences composed of veridical inferences and arbitrary inference types. We find that current NLI models do not perform consistently well on transitivity inference tasks, suggesting that they lack the generalization capacity for drawing composite inferences from provided training examples. The data and code for our analysis are publicly available at https://github.com/verypluming/transitivity.

Takeshi Kojima, Itsuki Okimura, Yusuke Iwasawa et al.

Current decoder-based pre-trained language models (PLMs) successfully demonstrate multilingual capabilities. However, it is unclear how these models handle multilingualism. We analyze the neuron-level internal behavior of multilingual decoder-based PLMs, Specifically examining the existence of neurons that fire ``uniquely for each language'' within decoder-only multilingual PLMs. We analyze six languages: English, German, French, Spanish, Chinese, and Japanese, and show that language-specific neurons are unique, with a slight overlap (< 5%) between languages. These neurons are mainly distributed in the models' first and last few layers. This trend remains consistent across languages and models. Additionally, we tamper with less than 1% of the total neurons in each model during inference and demonstrate that tampering with a few language-specific neurons drastically changes the probability of target language occurrence in text generation.

Yosuke Mikami, Daiki Matsuoka, Hitomi Yanaka

Natural Language Inference (NLI) involving comparatives is challenging because it requires understanding quantities and comparative relations expressed by sentences. While some approaches leverage Large Language Models (LLMs), we focus on logic-based approaches grounded in compositional semantics, which are promising for robust handling of numerical and logical expressions. Previous studies along these lines have proposed logical inference systems for English comparatives. However, it has been pointed out that there are several morphological and semantic differences between Japanese and English comparatives. These differences make it difficult to apply such systems directly to Japanese comparatives. To address this gap, this study proposes ccg-jcomp, a logical inference system for Japanese comparatives based on compositional semantics. We evaluate the proposed system on a Japanese NLI dataset containing comparative expressions. We demonstrate the effectiveness of our system by comparing its accuracy with that of existing LLMs.

Ryoma Kumon, Hitomi Yanaka

The compositional generalization abilities of neural models have been sought after for human-like linguistic competence. The popular method to evaluate such abilities is to assess the models' input-output behavior. However, that does not reveal the internal mechanisms, and the underlying competence of such models in compositional generalization remains unclear. To address this problem, we explore the inner workings of a Transformer model by finding an existing subnetwork that contributes to the generalization performance and by performing causal analyses on how the model utilizes syntactic features. We find that the model depends on syntactic features to output the correct answer, but that the subnetwork with much better generalization performance than the whole model relies on a non-compositional algorithm in addition to the syntactic features. We also show that the subnetwork improves its generalization performance relatively slowly during the training compared to the in-distribution one, and the non-compositional solution is acquired in the early stages of the training.

Rongzhi Li, Hitomi Yanaka

Extending large language models to low-resource languages is essential for global accessibility, but training separate models per language is prohibitively expensive. Mixture-of-Experts (MoE) architectures address this by adding sparse language-specific parameters, but determining how many experts each layer needs remains an open question. Current approaches allocate experts based on layer-level similarity, yet language processing exhibits fine-grained specialization at individual neurons. We propose $\textbf{NeuronMoE}$, a method that analyzes language-specific neurons across all transformer components to guide expert allocation per layer based on empirically measured cross-lingual neuron diversity. Applied to Llama-3.2-3B for low-resource languages (Greek, Turkish, and Hungarian), this approach achieves approximately 40% average parameter reduction while matching the performance of the LayerMoE baseline. We find that low-resource language experts independently develop neuron specialization patterns mirroring the high-resource language, which are concentrated in early and late layers. This reveals potential universal architectural principles in how multilingual models organize linguistic knowledge.

Taisei Yamamoto, Ryoma Kumon, Danushka Bollegala et al.

Large language models (LLMs) exhibit social biases, prompting the development of various debiasing methods. However, debiasing methods may degrade the capabilities of LLMs. Previous research has evaluated the impact of bias mitigation primarily through tasks measuring general language understanding, which are often unrelated to social biases. In contrast, cultural commonsense is closely related to social biases, as both are rooted in social norms and values. The impact of bias mitigation on cultural commonsense in LLMs has not been well investigated. Considering this gap, we propose SOBACO (SOcial BiAs and Cultural cOmmonsense benchmark), a Japanese benchmark designed to evaluate social biases and cultural commonsense in LLMs in a unified format. We evaluate several LLMs on SOBACO to examine how debiasing methods affect cultural commonsense in LLMs. Our results reveal that the debiasing methods degrade the performance of the LLMs on the cultural commonsense task (up to 75% accuracy deterioration). These results highlight the importance of developing debiasing methods that consider the trade-off with cultural commonsense to improve fairness and utility of LLMs.

Yosuke Mikami, Daiki Matsuoka, Hitomi Yanaka

Large Language Models (LLMs) perform remarkably well in Natural Language Inference (NLI). However, NLI involving numerical and logical expressions remains challenging. Comparatives are a key linguistic phenomenon related to such inference, but the robustness of LLMs in handling them, especially in languages that are not dominant in the models' training data, such as Japanese, has not been sufficiently explored. To address this gap, we construct a Japanese NLI dataset that focuses on comparatives and evaluate various LLMs in zero-shot and few-shot settings. Our results show that the performance of the models is sensitive to the prompt formats in the zero-shot setting and influenced by the gold labels in the few-shot examples. The LLMs also struggle to handle linguistic phenomena unique to Japanese. Furthermore, we observe that prompts containing logical semantic representations help the models predict the correct labels for inference problems that they struggle to solve even with few-shot examples.

Amane Watahiki, Tomoki Doi, Taiga Shinozaki et al.

One of the main objectives in developing large vision-language models (LVLMs) is to engineer systems that can assist humans with multimodal tasks, including interpreting descriptions of perceptual experiences. A central phenomenon in this context is amodal completion, in which people perceive objects even when parts of those objects are hidden. Although numerous studies have assessed whether computer-vision algorithms can detect or reconstruct occluded regions, the inferential abilities of LVLMs on texts related to amodal completion remain unexplored. To address this gap, we constructed a benchmark grounded in Basic Formal Ontology to achieve a systematic classification of amodal completion. Our results indicate that while many LVLMs achieve human-comparable performance overall, their accuracy diverges for certain types of objects being completed. Notably, in certain categories, some LLaVA-NeXT variants and Claude 3.5 Sonnet exhibit lower accuracy on original images compared to blank stimuli lacking visual content. Intriguingly, this disparity emerges only under Japanese prompting, suggesting a deficiency in Japanese-specific linguistic competence among these models.

Taiga Someya, Ryo Yoshida, Hitomi Yanaka et al.

Recent work has demonstrated that neural language models encode syntactic structures in their internal representations, yet the derivations by which these structures are constructed across layers remain poorly understood. In this paper, we propose Derivational Probing to investigate how micro-syntactic structures (e.g., subject noun phrases) and macro-syntactic structures (e.g., the relationship between the root verbs and their direct dependents) are constructed as word embeddings propagate upward across layers. Our experiments on BERT reveal a clear bottom-up derivation: micro-syntactic structures emerge in lower layers and are gradually integrated into a coherent macro-syntactic structure in higher layers. Furthermore, a targeted evaluation on subject-verb number agreement shows that the timing of constructing macro-syntactic structures is critical for downstream performance, suggesting an optimal timing for integrating global syntactic information.

Hitomi Yanaka, Xinqi He, Jie Lu et al.

An increasing number of studies have examined the social bias of rapidly developed large language models (LLMs). Although most of these studies have focused on bias occurring in a single social attribute, research in social science has shown that social bias often occurs in the form of intersectionality -- the constitutive and contextualized perspective on bias aroused by social attributes. In this study, we construct the Japanese benchmark inter-JBBQ, designed to evaluate the intersectional bias in LLMs on the question-answering setting. Using inter-JBBQ to analyze GPT-4o and Swallow, we find that biased output varies according to its contexts even with the equal combination of social attributes.

Ryoma Kumon, Daiki Matsuoka, Hitomi Yanaka

Compositional generalization refers to the ability to generalize to novel combinations of previously observed words and syntactic structures. Since it is regarded as a desired property of neural models, recent work has assessed compositional generalization in machine translation as well as semantic parsing. However, previous evaluations with machine translation have focused mostly on lexical generalization (i.e., generalization to unseen combinations of known words). Thus, it remains unclear to what extent models can translate sentences that require structural generalization (i.e., generalization to different sorts of syntactic structures). To address this question, we construct SGET, a machine translation dataset covering various types of compositional generalization with control of words and sentence structures. We evaluate neural machine translation models on SGET and show that they struggle more in structural generalization than in lexical generalization. We also find different performance trends in semantic parsing and machine translation, which indicates the importance of evaluations across various tasks.

Rongzhi Li, Takeru Matsuda, Hitomi Yanaka

Word embeddings represent words as multidimensional real vectors, facilitating data analysis and processing, but are often challenging to interpret. Independent Component Analysis (ICA) creates clearer semantic axes by identifying independent key features. Previous research has shown ICA's potential to reveal universal semantic axes across languages. However, it lacked verification of the consistency of independent components within and across languages. We investigated the consistency of semantic axes in two ways: both within a single language and across multiple languages. We first probed into intra-language consistency, focusing on the reproducibility of axes by performing ICA multiple times and clustering the outcomes. Then, we statistically examined inter-language consistency by verifying those axes' correspondences using statistical tests. We newly applied statistical methods to establish a robust framework that ensures the reliability and universality of semantic axes.

Tomoki Doi, Masaru Isonuma, Hitomi Yanaka

Recent work utilizes Large Language Models (LLMs) for topic modeling, generating comprehensible topic labels for given documents. However, their performance has mainly been evaluated qualitatively, and there remains room for quantitative investigation of their capabilities. In this paper, we quantitatively evaluate LLMs from multiple perspectives: the quality of topics, the impact of LLM-specific concerns, such as hallucination and shortcuts for limited documents, and LLMs' controllability of topic categories via prompts. Our findings show that LLMs can identify coherent and diverse topics with few hallucinations but may take shortcuts by focusing only on parts of documents. We also found that their controllability is limited.

Riko Suzuki, Hitomi Yanaka, Koji Mineshima et al.

This paper introduces a new video-and-language dataset with human actions for multimodal logical inference, which focuses on intentional and aspectual expressions that describe dynamic human actions. The dataset consists of 200 videos, 5,554 action labels, and 1,942 action triplets of the form <subject, predicate, object> that can be translated into logical semantic representations. The dataset is expected to be useful for evaluating multimodal inference systems between videos and semantically complicated sentences including negation and quantification.

Masato Mita, Hitomi Yanaka

There has been an increased interest in data generation approaches to grammatical error correction (GEC) using pseudo data. However, these approaches suffer from several issues that make them inconvenient for real-world deployment including a demand for large amounts of training data. On the other hand, some errors based on grammatical rules may not necessarily require a large amount of data if GEC models can realize grammatical generalization. This study explores to what extent GEC models generalize grammatical knowledge required for correcting errors. We introduce an analysis method using synthetic and real GEC datasets with controlled vocabularies to evaluate whether models can generalize to unseen errors. We found that a current standard Transformer-based GEC model fails to realize grammatical generalization even in simple settings with limited vocabulary and syntax, suggesting that it lacks the generalization ability required to correct errors from provided training examples.

Hitomi Yanaka, Koji Mineshima, Daisuke Bekki et al.

Despite the success of language models using neural networks, it remains unclear to what extent neural models have the generalization ability to perform inferences. In this paper, we introduce a method for evaluating whether neural models can learn systematicity of monotonicity inference in natural language, namely, the regularity for performing arbitrary inferences with generalization on composition. We consider four aspects of monotonicity inferences and test whether the models can systematically interpret lexical and logical phenomena on different training/test splits. A series of experiments show that three neural models systematically draw inferences on unseen combinations of lexical and logical phenomena when the syntactic structures of the sentences are similar between the training and test sets. However, the performance of the models significantly decreases when the structures are slightly changed in the test set while retaining all vocabularies and constituents already appearing in the training set. This indicates that the generalization ability of neural models is limited to cases where the syntactic structures are nearly the same as those in the training set.

Hitomi Yanaka, Koji Mineshima, Daisuke Bekki et al.

Monotonicity reasoning is one of the important reasoning skills for any intelligent natural language inference (NLI) model in that it requires the ability to capture the interaction between lexical and syntactic structures. Since no test set has been developed for monotonicity reasoning with wide coverage, it is still unclear whether neural models can perform monotonicity reasoning in a proper way. To investigate this issue, we introduce the Monotonicity Entailment Dataset (MED). Performance by state-of-the-art NLI models on the new test set is substantially worse, under 55%, especially on downward reasoning. In addition, analysis using a monotonicity-driven data augmentation method showed that these models might be limited in their generalization ability in upward and downward reasoning.

Riko Suzuki, Hitomi Yanaka, Masashi Yoshikawa et al.

A large amount of research about multimodal inference across text and vision has been recently developed to obtain visually grounded word and sentence representations. In this paper, we use logic-based representations as unified meaning representations for texts and images and present an unsupervised multimodal logical inference system that can effectively prove entailment relations between them. We show that by combining semantic parsing and theorem proving, the system can handle semantically complex sentences for visual-textual inference.

Hitomi Yanaka, Koji Mineshima, Daisuke Bekki et al.

Large crowdsourced datasets are widely used for training and evaluating neural models on natural language inference (NLI). Despite these efforts, neural models have a hard time capturing logical inferences, including those licensed by phrase replacements, so-called monotonicity reasoning. Since no large dataset has been developed for monotonicity reasoning, it is still unclear whether the main obstacle is the size of datasets or the model architectures themselves. To investigate this issue, we introduce a new dataset, called HELP, for handling entailments with lexical and logical phenomena. We add it to training data for the state-of-the-art neural models and evaluate them on test sets for monotonicity phenomena. The results showed that our data augmentation improved the overall accuracy. We also find that the improvement is better on monotonicity inferences with lexical replacements than on downward inferences with disjunction and modification. This suggests that some types of inferences can be improved by our data augmentation while others are immune to it.

Hitomi Yanaka, Koji Mineshima, Pascual Martinez-Gomez et al.

How to identify, extract, and use phrasal knowledge is a crucial problem for the task of Recognizing Textual Entailment (RTE). To solve this problem, we propose a method for detecting paraphrases via natural deduction proofs of semantic relations between sentence pairs. Our solution relies on a graph reformulation of partial variable unifications and an algorithm that induces subgraph alignments between meaning representations. Experiments show that our method can automatically detect various paraphrases that are absent from existing paraphrase databases. In addition, the detection of paraphrases using proof information improves the accuracy of RTE tasks.

Hitomi Yanaka, Koji Mineshima, Pascual Martinez-Gomez et al.

Determining semantic textual similarity is a core research subject in natural language processing. Since vector-based models for sentence representation often use shallow information, capturing accurate semantics is difficult. By contrast, logical semantic representations capture deeper levels of sentence semantics, but their symbolic nature does not offer graded notions of textual similarity. We propose a method for determining semantic textual similarity by combining shallow features with features extracted from natural deduction proofs of bidirectional entailment relations between sentence pairs. For the natural deduction proofs, we use ccg2lambda, a higher-order automatic inference system, which converts Combinatory Categorial Grammar (CCG) derivation trees into semantic representations and conducts natural deduction proofs. Experiments show that our system was able to outperform other logic-based systems and that features derived from the proofs are effective for learning textual similarity.