Shuhei Kurita

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/.

Issa Sugiura, Shuhei Kurita, Yusuke Oda et al.

Understanding chart and table images is essential for applying vision-language models (VLMs) to real-world document understanding. While English benchmarks have advanced rapidly, non-English counterparts remain scarce, leaving it unclear whether this progress generalizes across languages. A key obstacle is the difficulty of collecting realistic and diverse non-English chart and table images at scale. To address this, we leverage governmental white papers as a scalable source for benchmark construction beyond English, as they contain naturally occurring charts and tables across diverse formats and domains and are freely accessible in many countries. As a first instantiation, we introduce HakushoBench, a challenging Japanese chart and table VQA benchmark built from 33 governmental white papers. HakushoBench contains 2,053 images spanning over 10 image types, with manually annotated QA pairs, designed to assess deep and holistic understanding of charts and tables, rather than local visual cues alone. Experiments across a broad range of VLMs demonstrate that HakushoBench remains challenging for open-weight models: the best open-weight model achieves only 58.6% accuracy, and a 34.9-point gap between open-weight and proprietary models highlights substantial room for improvement in complex chart and table understanding. We release our dataset and code.

Shuhei Kurita, Naoki Katsura, Eri Onami

Grounding textual expressions on scene objects from first-person views is a truly demanding capability in developing agents that are aware of their surroundings and behave following intuitive text instructions. Such capability is of necessity for glass-devices or autonomous robots to localize referred objects in the real-world. In the conventional referring expression comprehension tasks of images, however, datasets are mostly constructed based on the web-crawled data and don't reflect diverse real-world structures on the task of grounding textual expressions in diverse objects in the real world. Recently, a massive-scale egocentric video dataset of Ego4D was proposed. Ego4D covers around the world diverse real-world scenes including numerous indoor and outdoor situations such as shopping, cooking, walking, talking, manufacturing, etc. Based on egocentric videos of Ego4D, we constructed a broad coverage of the video-based referring expression comprehension dataset: RefEgo. Our dataset includes more than 12k video clips and 41 hours for video-based referring expression comprehension annotation. In experiments, we combine the state-of-the-art 2D referring expression comprehension models with the object tracking algorithm, achieving the video-wise referred object tracking even in difficult conditions: the referred object becomes out-of-frame in the middle of the video or multiple similar objects are presented in the video. Codes are available at https://github.com/shuheikurita/RefEgo

Taiki Miyanishi, Fumiya Kitamori, Shuhei Kurita et al.

City-scale 3D point cloud is a promising way to express detailed and complicated outdoor structures. It encompasses both the appearance and geometry features of segmented city components, including cars, streets, and buildings, that can be utilized for attractive applications such as user-interactive navigation of autonomous vehicles and drones. However, compared to the extensive text annotations available for images and indoor scenes, the scarcity of text annotations for outdoor scenes poses a significant challenge for achieving these applications. To tackle this problem, we introduce the CityRefer dataset for city-level visual grounding. The dataset consists of 35k natural language descriptions of 3D objects appearing in SensatUrban city scenes and 5k landmarks labels synchronizing with OpenStreetMap. To ensure the quality and accuracy of the dataset, all descriptions and labels in the CityRefer dataset are manually verified. We also have developed a baseline system that can learn encoded language descriptions, 3D object instances, and geographical information about the city's landmarks to perform visual grounding on the CityRefer dataset. To the best of our knowledge, the CityRefer dataset is the largest city-level visual grounding dataset for localizing specific 3D objects.

Koya Sakamoto, Taiki Miyanishi, Daichi Azuma et al.

Visual search in 3D environments requires embodied agents to actively explore their surroundings and acquire task-relevant evidence. However, existing visual search and embodied AI benchmarks, including EQA, typically rely on static observations or constrained egocentric motion, and thus do not explicitly evaluate fine-grained viewpoint-dependent phenomena that arise under unrestricted 5-DoF viewpoint control in real-world 3D environments, such as visibility changes caused by vertical viewpoint shifts, revealing contents inside containers, and disambiguating object attributes that are only observable from specific angles. To address this limitation, we introduce {E3VS-Bench}, a benchmark for embodied 3D visual search where agents must control their viewpoints in 5-DoF to gather viewpoint-dependent evidence for question answering. E3VS-Bench consists of 99 high-fidelity 3D scenes reconstructed using 3D Gaussian Splatting and 2,014 question-driven episodes. 3D Gaussian Splatting enables photorealistic free-viewpoint rendering that preserves fine-grained visual details (e.g., small text and subtle attributes) often degraded in mesh-based simulators, thereby allowing the construction of questions that cannot be answered from a single view and instead require active inspection across viewpoints in 5-DoF. We evaluate multiple state-of-the-art VLMs and compare their performance with humans. Despite strong 2D reasoning ability, all models exhibit a substantial gap from humans, highlighting limitations in active perception and coherent viewpoint planning specifically under full 5-DoF viewpoint changes.

Keisuke Shirai, Atsushi Hashimoto, Taichi Nishimura et al.

We present a new multimodal dataset called Visual Recipe Flow, which enables us to learn each cooking action result in a recipe text. The dataset consists of object state changes and the workflow of the recipe text. The state change is represented as an image pair, while the workflow is represented as a recipe flow graph (r-FG). The image pairs are grounded in the r-FG, which provides the cross-modal relation. With our dataset, one can try a range of applications, from multimodal commonsense reasoning and procedural text generation.

Daichi Azuma, Taiki Miyanishi, Shuhei Kurita et al.

In an era characterized by advancements in artificial intelligence and robotics, enabling machines to interact with and understand their environment is a critical research endeavor. In this paper, we propose Answerability Fields, a novel approach to predicting answerability within complex indoor environments. Leveraging a 3D question answering dataset, we construct a comprehensive Answerability Fields dataset, encompassing diverse scenes and questions from ScanNet. Using a diffusion model, we successfully infer and evaluate these Answerability Fields, demonstrating the importance of objects and their locations in answering questions within a scene. Our results showcase the efficacy of Answerability Fields in guiding scene-understanding tasks, laying the foundation for their application in enhancing interactions between intelligent agents and their environments.

Hisayuki Yokomizo, Taiki Miyanishi, Yan Gang et al.

Vision-Language Models (VLMs) are increasingly applied to robotic perception and manipulation, yet their ability to infer physical properties required for manipulation remains limited. In particular, estimating the mass of real-world objects is essential for determining appropriate grasp force and ensuring safe interaction. However, current VLMs lack reliable mass reasoning capabilities, and most existing benchmarks do not explicitly evaluate physical quantity estimation under realistic sensing conditions. In this work, we propose PhysQuantAgent, a framework for real-world object mass estimation using VLMs, together with VisPhysQuant, a new benchmark dataset for evaluation. VisPhysQuant consists of RGB-D videos of real objects captured from multiple viewpoints, annotated with precise mass measurements. To improve estimation accuracy, we introduce three visual prompting methods that enhance the input image with object detection, scale estimation, and cross-sectional image generation to help the model comprehend the size and internal structure of the target object. Experiments show that visual prompting significantly improves mass estimation accuracy on real-world data, suggesting the efficacy of integrating spatial reasoning with VLM knowledge for physical inference.

Mahiro Ukai, Shuhei Kurita, Atsushi Hashimoto et al.

Visual question answering aims to provide responses to natural language questions given visual input. Recently, visual programmatic models (VPMs), which generate executable programs to answer questions through large language models (LLMs), have attracted research interest. However, they often require long input prompts to provide the LLM with sufficient API usage details to generate relevant code. To address this limitation, we propose AdaCoder, an adaptive prompt compression framework for VPMs. AdaCoder operates in two phases: a compression phase and an inference phase. In the compression phase, given a preprompt that describes all API definitions in the Python language with example snippets of code, a set of compressed preprompts is generated, each depending on a specific question type. In the inference phase, given an input question, AdaCoder predicts the question type and chooses the appropriate corresponding compressed preprompt to generate code to answer the question. Notably, AdaCoder employs a single frozen LLM and pre-defined prompts, negating the necessity of additional training and maintaining adaptability across different powerful black-box LLMs such as GPT and Claude. In experiments, we apply AdaCoder to ViperGPT and demonstrate that it reduces token length by 71.1%, while maintaining or even improving the performance of visual question answering.

Kaito Baba, Chaoran Liu, Shuhei Kurita et al.

We present Prover Agent, a novel AI agent for automated theorem proving that integrates large language models (LLMs) with a formal proof assistant, Lean. Prover Agent coordinates an informal reasoning LLM, a formal prover model, and feedback from Lean while also generating auxiliary lemmas. These auxiliary lemmas are not limited to subgoals in the formal proof but can also include special cases or potentially useful facts derived from the assumptions, which help in discovering a viable proof strategy. It achieves an 88.1% success rate on the MiniF2F benchmark, establishing a new state-of-the-art among methods using small language models (SLMs) with a much lower sample budget than previous approaches. We also present theoretical analyses and case studies that illustrate how these generated lemmas contribute to solving challenging problems. Our code is publicly available at: https://github.com/kAIto47802/Prover-Agent.

Daichi Yashima, Shuhei Kurita, Yusuke Oda et al.

In this study, we focus on video captioning by fully open multimodal large language models (MLLMs). The comprehension of visual sequences is challenging because of their intricate temporal dependencies and substantial sequence length. The core attention mechanisms of existing Transformer-based approaches scale quadratically with the sequence length, making them computationally prohibitive. To address these limitations, we propose Aligned Hierarchical Bidirectional Scan Mamba (ABMamba), a fully open MLLM with linear computational complexity that enables the scalable processing of video sequences. ABMamba extends Deep State Space Models as its language backbone, replacing the costly quadratic attention mechanisms, and employs a novel Aligned Hierarchical Bidirectional Scan module that processes videos across multiple temporal resolutions. On standard video captioning benchmarks such as VATEX and MSR-VTT, ABMamba demonstrates competitive performance compared to typical MLLMs while achieving approximately three times higher throughput.

Daichi Yashima, Koki Seno, Shuhei Kurita et al.

Coarse-to-fine autoregressive modeling has recently shown strong promise for visuomotor policy learning, combining the inference efficiency of autoregressive methods with the global trajectory coherence of diffusion-based policies. However, existing approaches rely on discrete action tokenizers that map continuous action sequences to codebook indices, a design inherited from image generation where learned compression is necessary for high-dimensional pixel data. We observe that robot actions are inherently low-dimensional continuous vectors, for which such tokenization introduces unnecessary quantization error and a multi-stage training pipeline. In this work, we propose Hierarchical Flow Policy (HiFlow), a tokenization-free coarse-to-fine autoregressive policy that operates directly on raw continuous actions. HiFlow constructs multi-scale continuous action targets from each action chunk via simple temporal pooling. Specifically, it averages contiguous action windows to produce coarse summaries that are refined at finer temporal resolutions. The entire model is trained end-to-end in a single stage, eliminating the need for a separate tokenizer. Experiments on MimicGen, RoboTwin 2.0, and real-world environments demonstrate that HiFlow consistently outperforms existing methods including diffusion-based and tokenization-based autoregressive policies.

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.

Shunsuke Yasuki, Taiki Miyanishi, Nakamasa Inoue et al.

The advancement of 3D language fields has enabled intuitive interactions with 3D scenes via natural language. However, existing approaches are typically limited to small-scale environments, lacking the scalability and compositional reasoning capabilities necessary for large, complex urban settings. To overcome these limitations, we propose GeoProg3D, a visual programming framework that enables natural language-driven interactions with city-scale high-fidelity 3D scenes. GeoProg3D consists of two key components: (i) a Geography-aware City-scale 3D Language Field (GCLF) that leverages a memory-efficient hierarchical 3D model to handle large-scale data, integrated with geographic information for efficiently filtering vast urban spaces using directional cues, distance measurements, elevation data, and landmark references; and (ii) Geographical Vision APIs (GV-APIs), specialized geographic vision tools such as area segmentation and object detection. Our framework employs large language models (LLMs) as reasoning engines to dynamically combine GV-APIs and operate GCLF, effectively supporting diverse geographic vision tasks. To assess performance in city-scale reasoning, we introduce GeoEval3D, a comprehensive benchmark dataset containing 952 query-answer pairs across five challenging tasks: grounding, spatial reasoning, comparison, counting, and measurement. Experiments demonstrate that GeoProg3D significantly outperforms existing 3D language fields and vision-language models across multiple tasks. To our knowledge, GeoProg3D is the first framework enabling compositional geographic reasoning in high-fidelity city-scale 3D environments via natural language. The code is available at https://snskysk.github.io/GeoProg3D/.

Daichi Yashima, Shuhei Kurita, Yusuke Oda et al.

While multimodal large language models (MLLMs) have shown remarkable success across a wide range of tasks, long-form video understanding remains a significant challenge. In this study, we focus on video understanding by MLLMs. This task is challenging because processing a full stream of RGB frames is computationally intractable and highly redundant, as self-attention have quadratic complexity with sequence length. In this paper, we propose ReMoRa, a video MLLM that processes videos by operating directly on their compressed representations. A sparse set of RGB keyframes is retained for appearance, while temporal dynamics are encoded as a motion representation, removing the need for sequential RGB frames. These motion representations act as a compact proxy for optical flow, capturing temporal dynamics without full frame decoding. To refine the noise and low fidelity of block-based motions, we introduce a module to denoise and generate a fine-grained motion representation. Furthermore, our model compresses these features in a way that scales linearly with sequence length. We demonstrate the effectiveness of ReMoRa through extensive experiments across a comprehensive suite of long-video understanding benchmarks. ReMoRa outperformed baseline methods on multiple challenging benchmarks, including LongVideoBench, NExT-QA, and MLVU.

Eri Onami, Shuhei Kurita, Taiki Miyanishi et al.

Document question answering is a task of question answering on given documents such as reports, slides, pamphlets, and websites, and it is a truly demanding task as paper and electronic forms of documents are so common in our society. This is known as a quite challenging task because it requires not only text understanding but also understanding of figures and tables, and hence visual question answering (VQA) methods are often examined in addition to textual approaches. We introduce Japanese Document Question Answering (JDocQA), a large-scale document-based QA dataset, essentially requiring both visual and textual information to answer questions, which comprises 5,504 documents in PDF format and annotated 11,600 question-and-answer instances in Japanese. Each QA instance includes references to the document pages and bounding boxes for the answer clues. We incorporate multiple categories of questions and unanswerable questions from the document for realistic question-answering applications. We empirically evaluate the effectiveness of our dataset with text-based large language models (LLMs) and multimodal models. Incorporating unanswerable questions in finetuning may contribute to harnessing the so-called hallucination generation.

Tomoya Yoshida, Shuhei Kurita, Taichi Nishimura et al.

Visual affordance learning is a key component for robots to understand how to interact with objects. Conventional approaches in this field rely on pre-defined objects and actions, falling short of capturing diverse interactions in realworld scenarios. The key idea of our approach is employing textual instruction, targeting various affordances for a wide range of objects. This approach covers both hand-object and tool-object interactions. We introduce text-driven affordance learning, aiming to learn contact points and manipulation trajectories from an egocentric view following textual instruction. In our task, contact points are represented as heatmaps, and the manipulation trajectory as sequences of coordinates that incorporate both linear and rotational movements for various manipulations. However, when we gather data for this task, manual annotations of these diverse interactions are costly. To this end, we propose a pseudo dataset creation pipeline and build a large pseudo-training dataset: TextAFF80K, consisting of over 80K instances of the contact points, trajectories, images, and text tuples. We extend existing referring expression comprehension models for our task, and experimental results show that our approach robustly handles multiple affordances, serving as a new standard for affordance learning in real-world scenarios.

Koki Maeda, Shuhei Kurita, Taiki Miyanishi et al.

Given the accelerating progress of vision and language modeling, accurate evaluation of machine-generated image captions remains critical. In order to evaluate captions more closely to human preferences, metrics need to discriminate between captions of varying quality and content. However, conventional metrics fail short of comparing beyond superficial matches of words or embedding similarities; thus, they still need improvement. This paper presents VisCE$^2$, a vision language model-based caption evaluation method. Our method focuses on visual context, which refers to the detailed content of images, including objects, attributes, and relationships. By extracting and organizing them into a structured format, we replace the human-written references with visual contexts and help VLMs better understand the image, enhancing evaluation performance. Through meta-evaluation on multiple datasets, we validated that VisCE$^2$ outperforms the conventional pre-trained metrics in capturing caption quality and demonstrates superior consistency with human judgment.

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.

Fumihiko Tsuchiya, Taiki Miyanishi, Mahiro Ukai et al.

Counting in long videos remains a fundamental yet underexplored challenge in computer vision. Real-world recordings often span tens of minutes or longer and contain sparse, diverse events, making long-range temporal reasoning particularly difficult. However, most existing video counting benchmarks focus on short clips and evaluate only the final numerical answer, providing little insight into what should be counted or whether models consistently identify relevant instances across time. We introduce EC-Bench, a benchmark that jointly evaluates enumeration, counting, and temporal evidence grounding in long-form videos. EC-Bench contains 152 videos longer than 30 minutes and 1,699 queries paired with explicit evidence spans. Across 22 multimodal large language models (MLLMs), the best model achieves only 29.98% accuracy on Enumeration and 23.74% on Counting, while human performance reaches 78.57% and 82.97%, respectively. Our analysis reveals strong relationships between enumeration accuracy, temporal grounding, and counting performance. These results highlight fundamental limitations of current MLLMs and establish EC-Bench as a challenging benchmark for long-form quantitative video reasoning.

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.

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.

Tomoya Yoshida, Shuhei Kurita, Taichi Nishimura et al.

Egocentric videos capture how humans manipulate objects and tools, providing diverse motion cues for learning object manipulation. Unlike the costly, expert-driven manual teleoperation commonly used in training Vision-Language-Action models (VLAs), egocentric videos offer a scalable alternative. However, prior studies that leverage such videos for training robot policies typically rely on auxiliary annotations, such as detailed hand-pose recordings. Consequently, it remains unclear whether VLAs can be trained directly from raw egocentric videos. In this work, we address this challenge by leveraging EgoScaler, a framework that extracts 6DoF object manipulation trajectories from egocentric videos without requiring auxiliary recordings. We apply EgoScaler to four large-scale egocentric video datasets and automatically refine noisy or incomplete trajectories, thereby constructing a new large-scale dataset for VLA pre-training. Our experiments with a state-of-the-art $π_0$ architecture in both simulated and real-robot environments yield three key findings: (i) pre-training on our dataset improves task success rates by over 20\% compared to training from scratch, (ii) the performance is competitive with that achieved using real-robot datasets, and (iii) combining our dataset with real-robot data yields further improvements. These results demonstrate that egocentric videos constitute a promising and scalable resource for advancing VLA research.

Tomoya Yoshida, Shuhei Kurita, Taichi Nishimura et al.

Learning to use tools or objects in common scenes, particularly handling them in various ways as instructed, is a key challenge for developing interactive robots. Training models to generate such manipulation trajectories requires a large and diverse collection of detailed manipulation demonstrations for various objects, which is nearly unfeasible to gather at scale. In this paper, we propose a framework that leverages large-scale ego- and exo-centric video datasets -- constructed globally with substantial effort -- of Exo-Ego4D to extract diverse manipulation trajectories at scale. From these extracted trajectories with the associated textual action description, we develop trajectory generation models based on visual and point cloud-based language models. In the recently proposed egocentric vision-based in-a-quality trajectory dataset of HOT3D, we confirmed that our models successfully generate valid object trajectories, establishing a training dataset and baseline models for the novel task of generating 6DoF manipulation trajectories from action descriptions in egocentric vision.

Eri Onami, Taiki Miyanishi, Koki Maeda et al.

Legal documents including judgments and court orders require highly sophisticated legal knowledge for understanding. To disclose expert knowledge for non-experts, we explore the problem of visualizing legal texts with easy-to-understand diagrams and propose a novel dataset of LegalViz with 23 languages and 7,010 cases of legal document and visualization pairs, using the DOT graph description language of Graphviz. LegalViz provides a simple diagram from a complicated legal corpus identifying legal entities, transactions, legal sources, and statements at a glance, that are essential in each judgment. In addition, we provide new evaluation metrics for the legal diagram visualization by considering graph structures, textual similarities, and legal contents. We conducted empirical studies on few-shot and finetuning large language models for generating legal diagrams and evaluated them with these metrics, including legal content-based evaluation within 23 languages. Models trained with LegalViz outperform existing models including GPTs, confirming the effectiveness of our dataset.

Mahiro Ukai, Shuhei Kurita, Nakamasa Inoue

Object state recognition aims to identify the specific condition of objects, such as their positional states (e.g., open or closed) and functional states (e.g., on or off). While recent Vision-Language Models (VLMs) are capable of performing a variety of multimodal tasks, it remains unclear how precisely they can identify object states. To alleviate this issue, we introduce the STAte and Transition UnderStanding Benchmark (STATUS Bench), the first benchmark for rigorously evaluating the ability of VLMs to understand subtle variations in object states in diverse situations. Specifically, STATUS Bench introduces a novel evaluation scheme that requires VLMs to perform three tasks simultaneously: object state identification (OSI), image retrieval (IR), and state change identification (SCI). These tasks are defined over our fully hand-crafted dataset involving image pairs, their corresponding object state descriptions and state change descriptions. Furthermore, we introduce a large-scale training dataset, namely STATUS Train, which consists of 13 million semi-automatically created descriptions. This dataset serves as the largest resource to facilitate further research in this area. In our experiments, we demonstrate that STATUS Bench enables rigorous consistency evaluation and reveal that current state-of-the-art VLMs still significantly struggle to capture subtle object state distinctions. Surprisingly, under the proposed rigorous evaluation scheme, most open-weight VLMs exhibited chance-level zero-shot performance. After fine-tuning on STATUS Train, Qwen2.5-VL achieved performance comparable to Gemini 2.0 Flash. These findings underscore the necessity of STATUS Bench and Train for advancing object state recognition in VLM research.

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.

Kanoko Goto, Takumi Hirose, Mahiro Ukai et al.

Referring expression comprehension (REC) aims to localize the target object described by a natural language expression. Recent advances in vision-language learning have led to significant performance improvements in REC tasks. However, localizing extremely small objects remains a considerable challenge despite its importance in real-world applications such as autonomous driving. To address this issue, we introduce a novel dataset and method for REC targeting small objects. First, we present the small object REC (SOREC) dataset, which consists of 100,000 pairs of referring expressions and corresponding bounding boxes for small objects in driving scenarios. Second, we propose the progressive-iterative zooming adapter (PIZA), an adapter module for parameter-efficient fine-tuning that enables models to progressively zoom in and localize small objects. In a series of experiments, we apply PIZA to GroundingDINO and demonstrate a significant improvement in accuracy on the SOREC dataset. Our dataset, codes and pre-trained models are publicly available on the project page.

Issa Sugiura, Shuhei Kurita, Yusuke Oda et al.

We propose Llama-Mimi, a speech language model that uses a unified tokenizer and a single Transformer decoder to jointly model sequences of interleaved semantic and acoustic tokens. Comprehensive evaluation shows that Llama-Mimi achieves state-of-the-art performance in acoustic consistency and possesses the ability to preserve speaker identity. Our analysis further demonstrates that increasing the number of quantizers improves acoustic fidelity but degrades linguistic performance, highlighting the inherent challenge of maintaining long-term coherence. We additionally introduce an LLM-as-a-Judge-based evaluation to assess the spoken content quality of generated outputs. Our models, code, and speech samples are publicly available.

Jungdae Lee, Taiki Miyanishi, Shuhei Kurita et al.

Vision-and-language navigation (VLN) aims to develop agents capable of navigating in realistic environments. While recent cross-modal training approaches have significantly improved navigation performance in both indoor and outdoor scenarios, aerial navigation over real-world cities remains underexplored primarily due to limited datasets and the difficulty of integrating visual and geographic information. To fill this gap, we introduce CityNav, the first large-scale real-world dataset for aerial VLN. Our dataset consists of 32,637 human demonstration trajectories, each paired with a natural language description, covering 4.65 km$^2$ across two real cities: Cambridge and Birmingham. In contrast to existing datasets composed of synthetic scenes such as AerialVLN, our dataset presents a unique challenge because agents must interpret spatial relationships between real-world landmarks and the navigation destination, making CityNav an essential benchmark for advancing aerial VLN. Furthermore, as an initial step toward addressing this challenge, we provide a methodology of creating geographic semantic maps that can be used as an auxiliary modality input during navigation. In our experiments, we compare performance of three representative aerial VLN agents (Seq2seq, CMA and AerialVLN models) and demonstrate that the semantic map representation significantly improves their navigation performance.

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.

Taiki Miyanishi, Daichi Azuma, Shuhei Kurita et al.

We present a novel task for cross-dataset visual grounding in 3D scenes (Cross3DVG), which overcomes limitations of existing 3D visual grounding models, specifically their restricted 3D resources and consequent tendencies of overfitting a specific 3D dataset. We created RIORefer, a large-scale 3D visual grounding dataset, to facilitate Cross3DVG. It includes more than 63k diverse descriptions of 3D objects within 1,380 indoor RGB-D scans from 3RScan, with human annotations. After training the Cross3DVG model using the source 3D visual grounding dataset, we evaluate it without target labels using the target dataset with, e.g., different sensors, 3D reconstruction methods, and language annotators. Comprehensive experiments are conducted using established visual grounding models and with CLIP-based multi-view 2D and 3D integration designed to bridge gaps among 3D datasets. For Cross3DVG tasks, (i) cross-dataset 3D visual grounding exhibits significantly worse performance than learning and evaluation with a single dataset because of the 3D data and language variants across datasets. Moreover, (ii) better object detector and localization modules and fusing 3D data and multi-view CLIP-based image features can alleviate this lower performance. Our Cross3DVG task can provide a benchmark for developing robust 3D visual grounding models to handle diverse 3D scenes while leveraging deep language understanding.

Daichi Azuma, Taiki Miyanishi, Shuhei Kurita et al.

We propose a new 3D spatial understanding task of 3D Question Answering (3D-QA). In the 3D-QA task, models receive visual information from the entire 3D scene of the rich RGB-D indoor scan and answer the given textual questions about the 3D scene. Unlike the 2D-question answering of VQA, the conventional 2D-QA models suffer from problems with spatial understanding of object alignment and directions and fail the object identification from the textual questions in 3D-QA. We propose a baseline model for 3D-QA, named ScanQA model, where the model learns a fused descriptor from 3D object proposals and encoded sentence embeddings. This learned descriptor correlates the language expressions with the underlying geometric features of the 3D scan and facilitates the regression of 3D bounding boxes to determine described objects in textual questions and outputs correct answers. We collected human-edited question-answer pairs with free-form answers that are grounded to 3D objects in each 3D scene. Our new ScanQA dataset contains over 40K question-answer pairs from the 800 indoor scenes drawn from the ScanNet dataset. To the best of our knowledge, the proposed 3D-QA task is the first large-scale effort to perform object-grounded question-answering in 3D environments.

Shuhei Kurita, Kyunghyun Cho

Vision-and-language navigation (VLN) is a task in which an agent is embodied in a realistic 3D environment and follows an instruction to reach the goal node. While most of the previous studies have built and investigated a discriminative approach, we notice that there are in fact two possible approaches to building such a VLN agent: discriminative \textit{and} generative. In this paper, we design and investigate a generative language-grounded policy which uses a language model to compute the distribution over all possible instructions i.e. all possible sequences of vocabulary tokens given action and the transition history. In experiments, we show that the proposed generative approach outperforms the discriminative approach in the Room-2-Room (R2R) and Room-4-Room (R4R) datasets, especially in the unseen environments. We further show that the combination of the generative and discriminative policies achieves close to the state-of-the art results in the R2R dataset, demonstrating that the generative and discriminative policies capture the different aspects of VLN.

Shuhei Kurita, Anders Søgaard

In Semantic Dependency Parsing (SDP), semantic relations form directed acyclic graphs, rather than trees. We propose a new iterative predicate selection (IPS) algorithm for SDP. Our IPS algorithm combines the graph-based and transition-based parsing approaches in order to handle multiple semantic head words. We train the IPS model using a combination of multi-task learning and task-specific policy gradient training. Trained this way, IPS achieves a new state of the art on the SemEval 2015 Task 18 datasets. Furthermore, we observe that policy gradient training learns an easy-first strategy.

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.