Ichiro Kobayashi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Ichiro Kobayashi has a total of 62 co-authors. Here are the top co-authors by number of shared publications:
- Yusuke Miyao
The University of Tokyo, Tokyo, Japan
12 shared publications - Hiroya Takamura
National Institute of Advanced Industrial Science and Technology
11 shared publications - Chung-Chi Chen
Human-Agent Ally Lab, National Institute of Informatics, Tokyo, Japan
5 shared publications - Masayuki Asahara
National Institute for Japanese Language and Linguistics
5 shared publications - Eri Kuroda
Ochanomizu University, 2-1-1 Ohtsuka, Bunkyo-ku, Tokyo 112-8610, Japan
4 shared publications - Fei Cheng
Kyoto University
3 shared publications
Latest Publications
- Open Access
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7007-7011
Can Frontier LLMs Capture $40 Billion in Supply Chain Value?
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 7007-7011
Open AccessCan Frontier LLMs Capture $40 Billion in Supply Chain Value?
- Open Access
- 20 July 2026
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 30 (4) , 1243-1257
fMRI Encoding and Decoding with LLMs: Input Embeddings vs. Hidden State Representations
Systematically comparing how linguistic representations relate to brain activity has become an important topic in the field of computational neuroscience. Prior studies have mainly relied on contextual hidden states combined with linear regression, leaving open questions about the role of static input embeddings and the benefits of nonlinear mappings. In this study, we compare input embeddings and hidden states from multiple language model families (BERT, GPT-2, and LLaMA) within both encoding frameworks, which map text features to brain responses, and decoding frameworks, which reconstruct linguistic features from brain activity. We benchmarked voxel-wise ridge regression against bidirectional long short-term memory (BiLSTMs) models, using repeat-split cross-validation and explainable variance normalization on functional magnetic resonance imaging (fMRI) data from three subjects. Our analyses demonstrate that input embeddings, despite being context-invariant, remain competitive and, in some cases, outperform hidden states, while BiLSTMs provide modest but region-specific improvements over ridge regression. Fine-grained voxel-level results further revealed distinct cortical distributions of stable versus context-dependent features. Together, these findings clarify the trade-off between predictive performance and interpretability and highlight that input embeddings offer a strong and interpretable baseline for representational alignment between language models and brain activity.
- 20 July 2026
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 30 (4) , 1243-1257
Open AccessfMRI Encoding and Decoding with LLMs: Input Embeddings vs. Hidden State Representations
Systematically comparing how linguistic representations relate to brain activity has become an important topic in the field of computational neuroscience. Prior studies have mainly relied on contextual hidden states combined with linear regression, leaving open questions about the role of static input embeddings and the benefits of nonlinear mappings. In this study, we compare input embeddings and hidden states from multiple language model families (BERT, GPT-2, and LLaMA) within both encoding frameworks, which map text features to brain responses, and decoding frameworks, which reconstruct linguistic features from brain activity. We benchmarked voxel-wise ridge regression against bidirectional long short-term memory (BiLSTMs) models, using repeat-split cross-validation and explainable variance normalization on functional magnetic resonance imaging (fMRI) data from three subjects. Our analyses demonstrate that input embeddings, despite being context-invariant, remain competitive and, in some cases, outperform hidden states, while BiLSTMs provide modest but region-specific improvements over ridge regression. Fine-grained voxel-level results further revealed distinct cortical distributions of stable versus context-dependent features. Together, these findings clarify the trade-off between predictive performance and interpretability and highlight that input embeddings offer a strong and interpretable baseline for representational alignment between language models and brain activity.
- 20 May 2026
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 30 (3) , 839-858
Emotion Analysis in the Human Brain Evoked by Language Stimulation
Advances in neural recording and deep learning have enabled a more precise analysis of how the brain processes language. However, neural representations of diverse emotions during natural narrative language comprehension remains unclear. In this study, we investigated how emotional content in naturalistic spoken language is reflected in brain activity using two independent narrative functional magnetic resonance imaging datasets: Alice and Le Petit Prince. Participants listened to narrative readings and their blood-oxygenation-level-dependent responses were recorded. An emotion recognition model was developed by fine-tuning BERT using the GoEmotions dataset together with sentence-level annotations for 80 emotions. Emotional and linguistic features were extracted and used to train voxel-wise encoding models to predict brain activity. To examine emotion-related neural representations, we subtracted predictions based on linguistic features from those based on combined emotional and linguistic features. The resulting emotion-related activity patterns were distributed across cortical regions associated with affective and semantic processing and were consistent across subjects and datasets. These findings indicate that the emotional meaning conveyed through natural language is represented in a graded and distributed manner in the human brain, beyond linguistic content alone.
- 20 May 2026
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 30 (3) , 839-858
Emotion Analysis in the Human Brain Evoked by Language Stimulation
Advances in neural recording and deep learning have enabled a more precise analysis of how the brain processes language. However, neural representations of diverse emotions during natural narrative language comprehension remains unclear. In this study, we investigated how emotional content in naturalistic spoken language is reflected in brain activity using two independent narrative functional magnetic resonance imaging datasets: Alice and Le Petit Prince. Participants listened to narrative readings and their blood-oxygenation-level-dependent responses were recorded. An emotion recognition model was developed by fine-tuning BERT using the GoEmotions dataset together with sentence-level annotations for 80 emotions. Emotional and linguistic features were extracted and used to train voxel-wise encoding models to predict brain activity. To examine emotion-related neural representations, we subtracted predictions based on linguistic features from those based on combined emotional and linguistic features. The resulting emotion-related activity patterns were distributed across cortical regions associated with affective and semantic processing and were consistent across subjects and datasets. These findings indicate that the emotional meaning conveyed through natural language is represented in a graded and distributed manner in the human brain, beyond linguistic content alone.
- 21 April 2026
- Published by Springer Nature
- p. 464-479
Enhancing LLM Abductive Reasoning Through MCMC Premise Retrieval
- 21 April 2026
- Published by Springer Nature
- p. 464-479
Enhancing LLM Abductive Reasoning Through MCMC Premise Retrieval
- 20 November 2025
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 29 (6) , 1417-1426
Interactive Image Caption Generation Reflecting User Intent from Trace Using a Diffusion Language Model
This study proposes an image captioning method designed to incorporate user-specific explanatory intentions into the generated text, as signaled by the user’s trace on the image. We extract areas of interest from dense sections of the trace, determine the order of explanations by tracking changes in the pen-tip coordinates, and assess the degree of interest in each area by analyzing the time spent on them. Additionally, a diffusion language model is utilized to generate sentences in a non-autoregressive manner, allowing control over sentence length based on the temporal data of the trace. In the actual caption generation task, the proposed method achieved higher string similarity than conventional methods, including autoregressive models, and successfully captured user intent from the trace and faithfully reflected it in the generated text.
- 20 November 2025
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 29 (6) , 1417-1426
Interactive Image Caption Generation Reflecting User Intent from Trace Using a Diffusion Language Model
This study proposes an image captioning method designed to incorporate user-specific explanatory intentions into the generated text, as signaled by the user’s trace on the image. We extract areas of interest from dense sections of the trace, determine the order of explanations by tracking changes in the pen-tip coordinates, and assess the degree of interest in each area by analyzing the time spent on them. Additionally, a diffusion language model is utilized to generate sentences in a non-autoregressive manner, allowing control over sentence length based on the temporal data of the trace. In the actual caption generation task, the proposed method achieved higher string similarity than conventional methods, including autoregressive models, and successfully captured user intent from the trace and faithfully reflected it in the generated text.
- 20 November 2025
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 29 (6) , 1427-1442
Generating Natural Language Sentences Explaining Trends and Relationships of Two Time-Series Data
We propose a method for generating natural language explanations that describe trends and relationships between two time-series data. To address this task, it is essential to analyze the dynamic behavior of both time series and generate textual explanations based on the analytical outcomes. We developed a model that extended the vanilla Transformer architecture to better capture the temporal features relevant to explanation generation. To train the model, we constructed a synthetic, domain-agnostic dataset that simulated time-series patterns and interactions. We conducted two experiments to evaluate the effectiveness of the proposed approach using the synthesized datasets. The first experiment focused on generating explanations for the time-series trends. The results demonstrated that our model could generate accurate and coherent explanations with high accuracy. The second experiment addressed more complex scenarios in which the model was required to answer questions regarding the relationship between two interacting time-series. Although the model initially struggled to achieve high accuracy in this task, we observed that step-by-step training significantly improved its performance. These findings highlight both the potential and current limitations of Transformer-based approaches for interpretable time-series analysis.
- 20 November 2025
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 29 (6) , 1427-1442
Generating Natural Language Sentences Explaining Trends and Relationships of Two Time-Series Data
We propose a method for generating natural language explanations that describe trends and relationships between two time-series data. To address this task, it is essential to analyze the dynamic behavior of both time series and generate textual explanations based on the analytical outcomes. We developed a model that extended the vanilla Transformer architecture to better capture the temporal features relevant to explanation generation. To train the model, we constructed a synthetic, domain-agnostic dataset that simulated time-series patterns and interactions. We conducted two experiments to evaluate the effectiveness of the proposed approach using the synthesized datasets. The first experiment focused on generating explanations for the time-series trends. The results demonstrated that our model could generate accurate and coherent explanations with high accuracy. The second experiment addressed more complex scenarios in which the model was required to answer questions regarding the relationship between two interacting time-series. Although the model initially struggled to achieve high accuracy in this task, we observed that step-by-step training significantly improved its performance. These findings highlight both the potential and current limitations of Transformer-based approaches for interpretable time-series analysis.
- 9 November 2025
- Published by Springer Nature
- p. 384-398
Prompting Large Language Models for fMRI-Based Brain Semantic Decoding
- 9 November 2025
- Published by Springer Nature
- p. 384-398
Prompting Large Language Models for fMRI-Based Brain Semantic Decoding
- Open Access
- 20 September 2025
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 29 (5) , 1190-1202
Verbal Representation of Object Collision Prediction Based on Physical Properties
In recent years, prediction models for the real world have been widely proposed. Most research that deals with the recognition and prediction of the real world generates prediction results from visual predictions such as changes in pixels or numerical changes in physical simulators, and few models can predict them based on both visual and physical characteristics, as humans can. Therefore, in this study, we constructed a new prediction model based on both visual information and physical characteristics in the environment by integrating the mechanism of variational temporal abstraction, which extracts change points in the observation environment from visual information, into PreCNet. Furthermore, to make the prediction results interpretable, we generated the inferred prediction content as a sentence. In addition, we verified whether the generated sentences could explain collision situations in as much detail as a human being when given physical common sense about the environment, such as the movement and mass of objects.
- 20 September 2025
- Published by Fuji Technology Press Ltd. in Journal of Advanced Computational Intelligence and Intelligent Informatics
- Vol. 29 (5) , 1190-1202
Open AccessVerbal Representation of Object Collision Prediction Based on Physical Properties
In recent years, prediction models for the real world have been widely proposed. Most research that deals with the recognition and prediction of the real world generates prediction results from visual predictions such as changes in pixels or numerical changes in physical simulators, and few models can predict them based on both visual and physical characteristics, as humans can. Therefore, in this study, we constructed a new prediction model based on both visual information and physical characteristics in the environment by integrating the mechanism of variational temporal abstraction, which extracts change points in the observation environment from visual information, into PreCNet. Furthermore, to make the prediction results interpretable, we generated the inferred prediction content as a sentence. In addition, we verified whether the generated sentences could explain collision situations in as much detail as a human being when given physical common sense about the environment, such as the movement and mass of objects.
- 16 September 2025
- Published by Springer Nature in Neural Computing and Applications
- p. 1-18
Toward transfer learning integrating multiple functions through the latent space
- 16 September 2025
- Published by Springer Nature in Neural Computing and Applications
- p. 1-18
Toward transfer learning integrating multiple functions through the latent space