Daisuke Kawahara

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/.

Tomohito Kasahara, Daisuke Kawahara, Nguyen Tung et al.

Dialogue systems without consistent responses are not fascinating. In this study, we build a dialogue system that can respond based on a given character setting (persona) to bring consistency. Considering the trend of the rapidly increasing scale of language models, we propose an approach that uses prompt-tuning, which has low learning costs, on pre-trained large-scale language models. The results of automatic and manual evaluations in English and Japanese show that it is possible to build a dialogue system with more natural and personalized responses using less computational resources than fine-tuning.

Ritvik Choudhary, Daisuke Kawahara

Building open-domain dialogue systems capable of rich human-like conversational ability is one of the fundamental challenges in language generation. However, even with recent advancements in the field, existing open-domain generative models fail to capture and utilize external knowledge, leading to repetitive or generic responses to unseen utterances. Current work on knowledge-grounded dialogue generation primarily focuses on persona incorporation or searching a fact-based structured knowledge source such as Wikipedia. Our method takes a broader and simpler approach, which aims to improve the raw conversation ability of the system by mimicking the human response behavior through casual interactions found on social media. Utilizing a joint retriever-generator setup, the model queries a large set of filtered comment data from Reddit to act as additional context for the seq2seq generator. Automatic and human evaluations on open-domain dialogue datasets demonstrate the effectiveness of our approach.

Ryoma Sakaeda, Daisuke Kawahara

We aim to overcome the lack of diversity in responses of current dialogue systems and to develop a dialogue system that is engaging as a conversational partner. We propose a generator-evaluator model that evaluates multiple responses generated by a response generator and selects the best response by an evaluator. By generating multiple responses, we obtain diverse responses. We conduct human evaluations to compare the output of the proposed system with that of a baseline system. The results of the human evaluations showed that the proposed system's responses were often judged to be better than the baseline system's, and indicated the effectiveness of the proposed method.

Tatsuya Ide, Daisuke Kawahara

In communication, a human would recognize the emotion of an interlocutor and respond with an appropriate emotion, such as empathy and comfort. Toward developing a dialogue system with such a human-like ability, we propose a method to build a dialogue corpus annotated with two kinds of emotions. We collect dialogues from Twitter and annotate each utterance with the emotion that a speaker put into the utterance (expressed emotion) and the emotion that a listener felt after listening to the utterance (experienced emotion). We built a dialogue corpus in Japanese using this method, and its statistical analysis revealed the differences between expressed and experienced emotions. We conducted experiments on recognition of the two kinds of emotions. The experimental results indicated the difficulty in recognizing experienced emotions and the effectiveness of multi-task learning of the two kinds of emotions. We hope that the constructed corpus will facilitate the study on emotion recognition in a dialogue and emotion-aware dialogue response generation.

Tatsuya Ide, Eiki Murata, Daisuke Kawahara et al.

Despite the remarkable progress in natural language understanding with pretrained Transformers, neural language models often do not handle commonsense knowledge well. Toward commonsense-aware models, there have been attempts to obtain knowledge, ranging from automatic acquisition to crowdsourcing. However, it is difficult to obtain a high-quality knowledge base at a low cost, especially from scratch. In this paper, we propose PHALM, a method of building a knowledge graph from scratch, by prompting both crowdworkers and a large language model (LLM). We used this method to build a Japanese event knowledge graph and trained Japanese commonsense generation models. Experimental results revealed the acceptability of the built graph and inferences generated by the trained models. We also report the difference in prompting humans and an LLM. Our code, data, and models are available at github.com/nlp-waseda/comet-atomic-ja.

Xuyang Cao, Qianying Liu, Chuan Xiao et al.

In multilingual pretraining, the test loss of a pretrained model is heavily influenced by the proportion of each language in the pretraining data, namely the \textit{language mixture ratios}. Multilingual scaling laws can predict the test loss under different language mixture ratios and can therefore be used to estimate the optimal ratios. However, the current approaches to multilingual scaling laws do not measure the \textit{cross-lingual transfer} effect, resulting in suboptimal mixture ratios. In this paper, we consider multilingual pretraining as a cooperative game in which each language acts as a player that jointly contributes to pretraining, gaining the resulting reduction in test loss as the payoff. Consequently, from the perspective of cooperative game theory, we quantify the cross-lingual transfer from each language by its contribution in the game, and propose a game-theoretic multilingual scaling law called \textit{ShapleyLaw}. Our experiments show that ShapleyLaw outperforms baseline methods in model performance prediction and language mixture optimization.

Keito Sasagawa, Shuhei Kurita, Daisuke Kawahara

Multimodal Large Language Models (MLLMs) have seen rapid advances in recent years and are now being applied to visual document understanding tasks. They are expected to process a wide range of document images across languages, including Japanese. Understanding documents from images requires models to read what are written in them. Since some Japanese documents are written vertically, support for vertical writing is essential. However, research specifically focused on vertically written Japanese text remains limited. In this study, we evaluate the reading capability of existing MLLMs on vertically written Japanese text. First, we generate a synthetic Japanese OCR dataset by rendering Japanese texts into images, and use it for both model fine-tuning and evaluation. This dataset includes Japanese text in both horizontal and vertical writing. We also create an evaluation dataset sourced from the real-world document images containing vertically written Japanese text. Using these datasets, we demonstrate that the existing MLLMs perform worse on vertically written Japanese text than on horizontally written Japanese text. Furthermore, we show that training MLLMs on our synthesized Japanese OCR dataset results in improving the performance of models that previously could not handle vertical writing. The datasets and code are publicly available https://github.com/llm-jp/eval_vertical_ja.

Qianying Liu, Wenyu Guan, Sujian Li et al.

Automatically solving math word problems is a critical task in the field of natural language processing. Recent models have reached their performance bottleneck and require more high-quality data for training. We propose a novel data augmentation method that reverses the mathematical logic of math word problems to produce new high-quality math problems and introduce new knowledge points that can benefit learning the mathematical reasoning logic. We apply the augmented data on two SOTA math word problem solving models and compare our results with a strong data augmentation baseline. Experimental results show the effectiveness of our approach. We release our code and data at https://github.com/yiyunya/RODA.

Ranran Haoran Zhang, Qianying Liu, Aysa Xuemo Fan et al.

Joint entity and relation extraction aims to extract relation triplets from plain text directly. Prior work leverages Sequence-to-Sequence (Seq2Seq) models for triplet sequence generation. However, Seq2Seq enforces an unnecessary order on the unordered triplets and involves a large decoding length associated with error accumulation. These introduce exposure bias, which may cause the models overfit to the frequent label combination, thus deteriorating the generalization. We propose a novel Sequence-to-Unordered-Multi-Tree (Seq2UMTree) model to minimize the effects of exposure bias by limiting the decoding length to three within a triplet and removing the order among triplets. We evaluate our model on two datasets, DuIE and NYT, and systematically study how exposure bias alters the performance of Seq2Seq models. Experiments show that the state-of-the-art Seq2Seq model overfits to both datasets while Seq2UMTree shows significantly better generalization. Our code is available at https://github.com/WindChimeRan/OpenJERE .

Tomohito Kasahara, Daisuke Kawahara

Automatic evaluation of text generation is essential for improving the accuracy of generation tasks. In light of the current trend towards increasingly larger decoder-based language models, we investigate automatic evaluation methods based on such models for text generation. This paper compares various methods, including tuning with encoder-based models and large language models under equal conditions, on two different tasks, machine translation evaluation and semantic textual similarity, in two languages, Japanese and English. Experimental results show that compared to the tuned encoder-based models, the tuned decoder-based models perform poorly. The analysis of the causes for this suggests that the decoder-based models focus on surface word sequences and do not capture meaning. It is also revealed that in-context learning of very large decoder-based models such as ChatGPT makes it difficult to identify fine-grained semantic differences.

Hao Wang, Yiqun Sun, Pengfei Wei et al.

Vision-language models (VLMs) have advanced rapidly and are increasingly deployed in real-world applications, especially with the rise of agent-based systems. However, their safety has received relatively limited attention. Even the latest proprietary and open-weight VLMs remain highly vulnerable to adversarial attacks, leaving downstream applications exposed to significant risks. In this work, we propose a novel and lightweight adversarial attack detection framework based on sparse autoencoders (SAEs), termed SAEgis. By inserting an SAE module into a pretrained VLM and training it with standard reconstruction objectives, we find that the learned sparse latent features naturally capture attack-relevant signals. These features enable reliable classification of whether an input image has been adversarially perturbed, even for previously unseen samples. Extensive experiments show that SAEgis achieves strong performance across in-domain, cross-domain, and cross-attack settings, with particularly large improvements in cross-domain generalization compared to existing baselines. In addition, combining signals from multiple layers further improves robustness and stability. To the best of our knowledge, this is the first work to explore SAE as a plug-and-play mechanism for adversarial attack detection in VLMs. Our method requires no additional adversarial training, introduces minimal overhead, and provides a practical approach for improving the safety of real-world VLM systems.

Ziqi Yin, Hao Wang, Kaito Horio et al.

We investigate the impact of politeness levels in prompts on the performance of large language models (LLMs). Polite language in human communications often garners more compliance and effectiveness, while rudeness can cause aversion, impacting response quality. We consider that LLMs mirror human communication traits, suggesting they align with human cultural norms. We assess the impact of politeness in prompts on LLMs across English, Chinese, and Japanese tasks. We observed that impolite prompts often result in poor performance, but overly polite language does not guarantee better outcomes. The best politeness level is different according to the language. This phenomenon suggests that LLMs not only reflect human behavior but are also influenced by language, particularly in different cultural contexts. Our findings highlight the need to factor in politeness for cross-cultural natural language processing and LLM usage.

Issa Sugiura, Koki Maeda, Shuhei Kurita et al.

Reliable evaluation is essential for the development of vision-language models (VLMs). However, Japanese VQA benchmarks have undergone far less iterative refinement than their English counterparts. As a result, many existing benchmarks contain issues such as ambiguous questions, incorrect answers, and instances that can be solved without visual grounding, undermining evaluation reliability and leading to misleading conclusions in model comparisons. To address these limitations, we introduce JAMMEval, a refined collection of Japanese benchmarks for reliable VLM evaluation. It is constructed by systematically refining seven existing Japanese benchmark datasets through two rounds of human annotation, improving both data quality and evaluation reliability. In our experiments, we evaluate open-weight and proprietary VLMs on JAMMEval and analyze the capabilities of recent models on Japanese VQA. We further demonstrate the effectiveness of our refinement by showing that the resulting benchmarks yield evaluation scores that better reflect model capability, exhibit lower run-to-run variance, and improve the ability to distinguish between models of different capability levels. We release our dataset and code to advance reliable evaluation of VLMs.

Keito Sasagawa, Koki Maeda, Issa Sugiura et al.

To develop high-performing Visual Language Models (VLMs), it is essential to prepare multimodal resources, such as image-text pairs, interleaved data, and instruction data. While multimodal resources for English are abundant, there is a significant lack of corresponding resources for non-English languages, such as Japanese. To address this problem, we take Japanese as a non-English language and propose a method for rapidly creating Japanese multimodal datasets from scratch. We collect Japanese image-text pairs and interleaved data from web archives and generate Japanese instruction data directly from images using an existing VLM. Our experimental results show that a VLM trained on these native datasets outperforms those relying on machine-translated content.

Hao Wang, Pinzhi Huang, Jihan Yang et al.

The rapid evolution of multimodal large language models (MLLMs) has significantly enhanced their real-world applications. However, achieving consistent performance across languages, especially when integrating cultural knowledge, remains a significant challenge. To better assess this issue, we introduce two new benchmarks: KnowRecall and VisRecall, which evaluate cross-lingual consistency in MLLMs. KnowRecall is a visual question answering benchmark designed to measure factual knowledge consistency in 15 languages, focusing on cultural and historical questions about global landmarks. VisRecall assesses visual memory consistency by asking models to describe landmark appearances in 9 languages without access to images. Experimental results reveal that state-of-the-art MLLMs, including proprietary ones, still struggle to achieve cross-lingual consistency. This underscores the need for more robust approaches that produce truly multilingual and culturally aware models.

Issa Sugiura, Keito Sasagawa, Keisuke Nakao et al.

Developing vision-language models (VLMs) that generalize across diverse tasks requires large-scale training datasets with diverse content. In English, such datasets are typically constructed by aggregating and curating numerous existing visual question answering (VQA) resources. However, this strategy does not readily extend to other languages, where VQA datasets remain limited in both scale and domain coverage, posing a major obstacle to building high-quality multilingual and non-English VLMs. In this work, we introduce Jagle, the largest Japanese multimodal post-training dataset to date, comprising approximately 9.2 million instances across diverse tasks. Rather than relying on existing VQA datasets, we collect heterogeneous source data, including images, image-text pairs, and PDF documents, and generate VQA pairs through multiple strategies such as VLM-based QA generation, translation, and text rendering. Experiments demonstrate that a 2.2B model trained with Jagle achieves strong performance on Japanese tasks, surpassing InternVL3.5-2B in average score across ten Japanese evaluation tasks and approaching within five points of Qwen3-VL-2B-Instruct. Furthermore, combining Jagle with FineVision does not degrade English performance; instead, it improves English performance compared to training with FineVision alone. To facilitate reproducibility and future research, we release the dataset, trained models, and code.

Issa Sugiura, Shuhei Kurita, Yusuke Oda et al.

Large-scale and high-quality image-text pair datasets play an important role in developing high-performing Vision-Language Models (VLMs). In this work, we introduce WAON, a large-scale and high-quality Japanese image-text pair dataset containing approximately 155 million examples, collected from Common Crawl. Our dataset construction pipeline employs various techniques, including filtering and deduplication, which have been shown to be effective in previous studies. To evaluate its effectiveness, we also construct WAON-Bench, a manually curated benchmark for Japanese cultural image classification, consisting of 374 classes. To assess the effectiveness of our dataset, we conduct experiments using both WAON and the Japanese subset of ReLAION, one of the most widely used vision-language datasets. We fine-tune SigLIP2, a strong multilingual model, on both datasets. The results demonstrate that WAON enhances model performance on WAON-Bench more efficiently than ReLAION and achieves higher accuracy across all evaluated benchmarks. Furthermore, the model fine-tuned on WAON achieves state-of-the-art performance on several Japanese cultural benchmarks. We release our dataset, model, and code at https://speed1313.github.io/WAON.

Hao Wang, Eiki Murata, Lingfang Zhang et al.

Recent advances in multimodal large language models (MLLMs) have significantly enhanced video understanding capabilities, opening new possibilities for practical applications. Yet current video benchmarks focus largely on indoor scenes or short-range outdoor activities, leaving the challenges associated with long-distance travel largely unexplored. Mastering extended geospatial-temporal trajectories is critical for next-generation MLLMs, underpinning real-world tasks such as embodied-AI planning and navigation. To bridge this gap, we present VIR-Bench, a novel benchmark consisting of 200 travel videos that frames itinerary reconstruction as a challenging task designed to evaluate and push forward MLLMs' geospatial-temporal intelligence. Experimental results reveal that state-of-the-art MLLMs, including proprietary ones, struggle to achieve high scores, underscoring the difficulty of handling videos that span extended spatial and temporal scales. Moreover, we conduct an in-depth case study in which we develop a prototype travel-planning agent that leverages the insights gained from VIR-Bench. The agent's markedly improved itinerary recommendations verify that our evaluation protocol not only benchmarks models effectively but also translates into concrete performance gains in user-facing applications.

Hao Wang, Tetsuro Morimura, Ukyo Honda et al.

Non-autoregressive (NAR) language models are known for their low latency in neural machine translation (NMT). However, a performance gap exists between NAR and autoregressive models due to the large decoding space and difficulty in capturing dependency between target words accurately. Compounding this, preparing appropriate training data for NAR models is a non-trivial task, often exacerbating exposure bias. To address these challenges, we apply reinforcement learning (RL) to Levenshtein Transformer, a representative edit-based NAR model, demonstrating that RL with self-generated data can enhance the performance of edit-based NAR models. We explore two RL approaches: stepwise reward maximization and episodic reward maximization. We discuss the respective pros and cons of these two approaches and empirically verify them. Moreover, we experimentally investigate the impact of temperature setting on performance, confirming the importance of proper temperature setting for NAR models' training.

Hao Wang, Shuhei Kurita, Shuichiro Shimizu et al.

Audio-visual speech recognition (AVSR) is a multimodal extension of automatic speech recognition (ASR), using video as a complement to audio. In AVSR, considerable efforts have been directed at datasets for facial features such as lip-readings, while they often fall short in evaluating the image comprehension capabilities in broader contexts. In this paper, we construct SlideAVSR, an AVSR dataset using scientific paper explanation videos. SlideAVSR provides a new benchmark where models transcribe speech utterances with texts on the slides on the presentation recordings. As technical terminologies that are frequent in paper explanations are notoriously challenging to transcribe without reference texts, our SlideAVSR dataset spotlights a new aspect of AVSR problems. As a simple yet effective baseline, we propose DocWhisper, an AVSR model that can refer to textual information from slides, and confirm its effectiveness on SlideAVSR.

Hao Wang, Hirofumi Shimizu, Daisuke Kawahara

Recent studies in natural language processing (NLP) have focused on modern languages and achieved state-of-the-art results in many tasks. Meanwhile, little attention has been paid to ancient texts and related tasks. Classical Chinese first came to Japan approximately 2,000 years ago. It was gradually adapted to a Japanese form called Kanbun-Kundoku (Kanbun) in Japanese reading and translating methods, which has significantly impacted Japanese literature. However, compared to the rich resources for ancient texts in mainland China, Kanbun resources remain scarce in Japan. To solve this problem, we construct the first Classical-Chinese-to-Kanbun dataset in the world. Furthermore, we introduce two tasks, character reordering and machine translation, both of which play a significant role in Kanbun comprehension. We also test the current language models on these tasks and discuss the best evaluation method by comparing the results with human scores. We release our code and dataset on GitHub.

Tatsuya Ide, Daisuke Kawahara

For a computer to naturally interact with a human, it needs to be human-like. In this paper, we propose a neural response generation model with multi-task learning of generation and classification, focusing on emotion. Our model based on BART (Lewis et al., 2020), a pre-trained transformer encoder-decoder model, is trained to generate responses and recognize emotions simultaneously. Furthermore, we weight the losses for the tasks to control the update of parameters. Automatic evaluations and crowdsourced manual evaluations show that the proposed model makes generated responses more emotionally aware.

Akiko Aizawa, Frederic Bergeron, Junjie Chen et al.

The global pandemic of COVID-19 has made the public pay close attention to related news, covering various domains, such as sanitation, treatment, and effects on education. Meanwhile, the COVID-19 condition is very different among the countries (e.g., policies and development of the epidemic), and thus citizens would be interested in news in foreign countries. We build a system for worldwide COVID-19 information aggregation containing reliable articles from 10 regions in 7 languages sorted by topics. Our reliable COVID-19 related website dataset collected through crowdsourcing ensures the quality of the articles. A neural machine translation module translates articles in other languages into Japanese and English. A BERT-based topic-classifier trained on our article-topic pair dataset helps users find their interested information efficiently by putting articles into different categories.

Shuhei Kurita, Daisuke Kawahara, Sadao Kurohashi

Japanese predicate-argument structure (PAS) analysis involves zero anaphora resolution, which is notoriously difficult. To improve the performance of Japanese PAS analysis, it is straightforward to increase the size of corpora annotated with PAS. However, since it is prohibitively expensive, it is promising to take advantage of a large amount of raw corpora. In this paper, we propose a novel Japanese PAS analysis model based on semi-supervised adversarial training with a raw corpus. In our experiments, our model outperforms existing state-of-the-art models for Japanese PAS analysis.

Abhishek Kumar, Daisuke Kawahara, Sadao Kurohashi

We propose a novel two-layered attention network based on Bidirectional Long Short-Term Memory for sentiment analysis. The novel two-layered attention network takes advantage of the external knowledge bases to improve the sentiment prediction. It uses the Knowledge Graph Embedding generated using the WordNet. We build our model by combining the two-layered attention network with the supervised model based on Support Vector Regression using a Multilayer Perceptron network for sentiment analysis. We evaluate our model on the benchmark dataset of SemEval 2017 Task 5. Experimental results show that the proposed model surpasses the top system of SemEval 2017 Task 5. The model performs significantly better by improving the state-of-the-art system at SemEval 2017 Task 5 by 1.7 and 3.7 points for sub-tracks 1 and 2 respectively.

Xun Wang, Katsuhito Sudoh, Masaaki Nagata et al.

This paper introduces a novel neural network model for question answering, the \emph{entity-based memory network}. It enhances neural networks' ability of representing and calculating information over a long period by keeping records of entities contained in text. The core component is a memory pool which comprises entities' states. These entities' states are continuously updated according to the input text. Questions with regard to the input text are used to search the memory pool for related entities and answers are further predicted based on the states of retrieved entities. Compared with previous memory network models, the proposed model is capable of handling fine-grained information and more sophisticated relations based on entities. We formulated several different tasks as question answering problems and tested the proposed model. Experiments reported satisfying results.