Our activities in Fiscal Year 2025 (from April 2025 to March 2026 — the tenth year) include educational and R&D businesses, implemented according to our policy for FY2025.
Educational Business
The goal of our educational business is to help people who conduct research on the WBA approach on a long-term basis. In FY2025, WBAI held the tenth WBA symposium (in Japanese) and the 3rd International Whole Brain Architecture Workshop, a tutorial, and a Special Session at the International Conference on Neural Information Processing (ICONIP2025) in November 2025.
The Tenth WBA Symposium
The symposium was held (in Japanese) online/onsite on September 4th, 2025 with the theme of “AGI, Brain, and Society: Can human-brain-morphic AGI be a bridge of ‘trust’ between humans and AI society?” with speakers Yutaka Matsuo (WBAI/University of Tokyo), Masahiko Osawa (Nihon University), Satoshi Kurihara (Keio University), Akiko Aizawa (National Institute of Informatics), Koichi Takahashi (WBAI), Taiyo Hamada (Araya Co., Ltd.), and Hiroshi Yamakawa.
It was organized by WBAI, supported by the MEXT project: “Deciphering and Manipulating Brain Dynamics Underlying Emergence of Behavioral Change: a Multidisciplinary Biology Approach.”
In the symposium, Meysam Hashemi (INS, INSERM, Aix-Marseille University, France) and Aran Nayebi (Carnegie Mellon University, USA) were conferred the WBA Incentive Awards, and awards for Meritorious Service were presented to Yumiko Nishimura for her contributions as the coordinator of the WBA seminar executive committee to organize events such as the symposium and WBA seminars and to Kazuya Gokita (Garm, Inc.) for providing us with the office space in particular.
Events at ICONIP 2025
We organized the 3rd International Whole Brain Architecture Workshop, a tutorial, and a Special Session in November 2025.
Please visit https://wba-initiative.org/en/27609/ for detailed information.
R&D Business
The goal of WBAI’s R&D business is to support research activities in the WBA approach. WBAI has been promoting research and development of brain-inspired AGI through Brain Reference Architecture (BRA)-driven development since FY 2018. BRA-driven development is a methodology for developing brain-morphic software that uses Brain Information Flow (BIF) diagrams, which represent the mesoscopic anatomical structure of the brain, Hypothetical Component Diagrams (HCDs), which describe the component capabilities consistent with the BIF, and Function Realization Graphs (FRGs) that define the functions of HCD components in a hierarchical manner. Recently, we have been engaging in automating the process of creating the aforementioned graphs using the large language models (LLMs). Below, projects receiving financial support from the Academic Transformation Area Research program “Deciphering and Manipulating Brain Dynamics Underlying Emergence of Behavioral Change: a Multidisciplinary Biology Approach ” are marked with (Behavioral Change). Projects supported by the Matsuo-Iwasawa Laboratory at the University of Tokyo are marked with (Matsuo-Iwasawa Lab).
The major achievements in FY2025 include the establishment of a method for optimizing the design efficiency of BRA-driven development, the systematic development of a list of brain-inspired AGI capabilities, the initiation of a system for the automated construction and evaluation of BRA data using LLM, and the expansion of BRA data for several brain regions. These advancements strengthened the methodological and technical foundation for the completion of the WBRA alpha version by the end of February 2027. During the spring of 2026, the development plan was reviewed and refined in preparation for the completion of the alpha version.
BIF Construction
As part of the development of the Brain Information Flow (BIF) diagrams, we developed a system (BDBRA) that extracts neural connections from neuroscience literature and created a provisional version of WholeBIF-RDB, which contains approximately 51,000 connections from about 130,000 papers (Behavioral Change).
Progress has been made in organizing BIF data spanning diverse brain regions. Regarding the cerebral cortex, Suzuki et al. [1][2] constructed and published a neocortical connectome that estimates forward/backward projections between human neocortical areas by integrating non-human primate data (Behavioral Change).
HCD/FRG Construction
Design Efficiency Based on Capability-Constrainedness and Systematization of the Motif Library
As core methodologies for HCD/FRG construction, significant progress has been made in improving design efficiency by utilizing the concept of capability-constrainedness and in systematizing the Motif library that supports it. Maruyama et al. [3] demonstrated that brain-inspired software design can be made more efficient by constraining the parameter space of neural circuit motifs using capability-constrainedness, and clarified the correspondence between capabilities and the parameter space based on the Wilson-Cowan model. Yamakawa et al. and Tawatsuji et al. proposed a design principle in which capability-constrainedness determines mechanism, and constructed a capability-constrained mechanism library and a Motif library for bidirectional design. These results were presented at the 48th Annual Meeting of the Japanese Neuroscience Society (July 2025, Niigata in Japanese), the 20th Annual Meeting of the Society for Hippocampus and Higher Brain Functions (September 2025 in Japanese), and the 3rd WBA International Workshop (held as part of ICONIP2025, November 2025) (Behavioral Changes).
In addition, a project led by Akshat Verma at Ritsumeikan University to implement a biologically constrained model (BCM) of the hippocampus has begun, and we are proceeding with the examination of implementations with high biological validity, ranging from spiking neurons to population-averaged models (Behavioral Change).
Development of an Ability List and Automatic FRG Construction
As part of the systematic development of a competency list for brain-inspired AGI, we systematically extracted and organized competencies common to both the CHC (Cattell-Horn-Carroll) theory and O*NET, and established design guidelines for a competency list bounded by the scope of the central nervous system’s involvement. We decomposed each competency into an ability decomposition graphs (ADGs) to clarify the input-output relationships of the competencies (Behavioral Change).
Following the completion of the FRG automatic evaluation technology (FRG Checker), we completed a prototype LLM prompt that automatically generates HCD/FRG candidates given BIF data and TLFs (top-level functions) (Behavioral Change).
Expansion of BRA Data across Diverse Brain Regions
Yamauchi et al. [4][5][6] have constructed BRA data for a neocortical microcircuit with TLF as the inferential function, and publication in Neuroscience Research (January 2026) has been confirmed (Matsuo-Iwasawa Lab).
Regarding the hippocampus and amygdala, Taniguchi et al. published BRA data for the hippocampus-amygdala complex titled “AF24Hippocampus-Amygdala”[7], while Tawatsuji published amygdala-related BRA data titled “YT24ITcAmy”[8]. Regarding BRA research on emotion and affect in the amygdala, a paper was submitted to Neural Networks. Concerning Canonical Isocortical Loops (CILs), progress was made in constructing HCD/FRG models based on the spiral hypothesis of neocortical-basal ganglia-thalamic circuits [9], and BRA data on habit formation in the basal ganglia was submitted to BRAES (BRA Evaluation System). As new BRA data, Inoue et al. [10] provided the BRA dataset “MI25Imagination” for imagination (behavioral change), and Ko, et al. [11] provided the BRA dataset “SK25LHbDepressionModel,” a depression model inspired by the lateral habenula, to the 3rd WBA International Workshop (Behavioral Change).
Construction of Development Environments
BRAES (BRA Evaluation System), the environment for submitting and reviewing BRA data, also underwent continuous functional enhancements, including the release of BRA Data Template v2.2 (which includes support for the Review Tool), updates to the error code list, and the addition of an automatic capability extraction feature (Behavioral Change).
The concept for BITS (Brain-morphic software Implementation & Testing System)— a system designed to verify whether brain-morphic software implemented through BRA-driven development is consistent with the brain’s structure and function— took shape, and discussions deepened in collaboration with the Motif Project (Matsuo-Iwasawa Lab).
Initiatives toward a World where AI Coexists in Harmony with Humanity
As a concrete research effort based on this direction, we have proposed the NeuroQuad framework, which is an integrated architecture consisting of four functional domains: the Interpretability Functional Domain (IFD), the Control and Monitoring Domain (CMD), the Mediated Communication Domain (MCD), and the Knowledge Preservation Domain (KPD). Based on BRA-driven development, it provides design guidelines for brain-inspired AGI to function as a cognitive bridge between humanity and superintelligence. As part of these results, Nakashima et al. published an interpretable anomaly detection method for a hippocampus-inspired spatial cognition model in Lecture Notes in Computer Science (Springer Nature Singapore) [12] (Behavioral Change).
Furthermore, as a value-theoretical foundation supporting the realization of AI in harmony with humanity, Yamakawa and Tawatsuji published the paper in the Journal of the Japanese Society for Artificial Intelligence (in Japanese) [13]. This paper distinguishes the value of human-brain-inspired AGI into “transitional value”—which may be overtaken by universal intelligence through technological progress—and “permanent value”—which can be realized only by human-brain-inspired AGI regardless of technological progress—and systematizes the latter based on two pillars: experiential understanding (PHE) and partial knowability of computational principles. This framework provides a unified positioning for application areas such as mediator functions, brain-based interpretability and controllability, the dynamic inheritance of culture, and medical applications, and offers a theoretical foundation for the safe design of brain-inspired AI, including NeuroQuad.
WBAI Activities and Volunteering
Events such as the symposium in FY2025 were primarily carried out by unpaid regular members and volunteers (excluding contracted staff at the secretariat). Additionally, for the organization of events, we received support from paid volunteers funded by Grant-in-Aid for Scientific Research and part-time staff funded by the WBAI budget.
Financial Statements for FY2025
The balance sheet and cash flow for FY2025 are presented below (Table 1 and Table 2).
The revenues were 0.51 million yen and the expenditure totaled 0.74 million yen for administration and 0.72 million yen for operation (total expenditure was 1.46 million yen to yield 0.95 million yen deficit).
Most of the revenues of WBAI came from supporting members. (As of March 2026, there were five supporting members consisting of enterprises and individuals.).
The subcontractor/outsourcing expenses for operation includes the cost of R&D, which was mostly carried out with the budget of exterior research institutions. For the awardees of the 2025 WBAI Incentive Awards, we paid a total 100,000 yen (JPY) of grants.
The expenses for secretariat personnel (shown in Subcontracting/outsourcing Expenses) were apportioned by 50% (50/50) for operating and administrative expenses. The remuneration was paid to an accountant office. The rent for the office was free of charge by courtesy of Garm, Inc.
Table 1: Balance Sheet (as of March 31, 2026)
|
Items |
Amount (JPY) |
||||
|---|---|---|---|---|---|
|
Ⅰ
|
Assets |
|
|||
|
1. |
Current Assets |
|
|||
|
Cash and Saving Account |
5,521,404 |
||||
|
Total Current Assets |
5,521,404 |
||||
|
Total Assets |
5,521,404 |
||||
|
Ⅱ
|
Liabilities |
|
|||
|
1. |
Current Liabilities |
80,000 |
|||
|
Withholding Taxes |
6,126 |
||||
|
Total Current Liabilities |
86,126 |
||||
|
Total Liabilities |
86,126 |
||||
|
Ⅲ
|
Net Assets |
|
|||
|
Retained Net Assets at the Beginning of Period |
6,390,045 |
||||
|
Net assets variation |
-954,767 |
||||
|
Total Net Assets |
5,435,278 |
||||
|
Total Liabilities and Net Assets |
5,521,404 |
||||
Table 2: Cash Flow
|
Items |
Amount (JPY) |
||||
|
Ⅰ
|
Recurring Revenues |
|
|||
|
１ |
Fees |
|
|||
|
Fees from Regular Members |
130,000 |
||||
|
Fees from Supporting Members |
320,000 |
||||
|
Total Fees |
450,000 |
||||
|
２ |
Other Revenues |
|
|||
|
Interest Income |
1,820 |
||||
|
Other Income |
57,200 |
||||
|
Total Recurring Revenues |
509,020 |
||||
|
Ⅱ
|
Ordinary Expenses |
|
|||
|
１ |
Operating Expenses |
|
|||
|
⑴ Personnel Expenses |
4,680 |
||||
|
⑵ Total Other Expenses |
717,902 |
||||
|
Meetings |
46,530 |
||||
|
Subcontractor/outsourcing Expenses |
473,000 |
||||
|
Communication |
94,822 |
||||
|
Prizes/Awards |
100,000 |
||||
|
Other |
3,550 |
||||
|
Total Operating Expenses |
722,582 |
||||
|
２ |
Administrative Expenses |
|
|||
|
⑴ Personnel Expenses |
0 |
||||
|
⑵ Total Other Expenses |
741,205 |
||||
|
Subcontractor/Outsourcing Expenses |
473,000 |
||||
|
Remuneration |
264,000 |
||||
|
Other |
4,205 |
||||
|
Total Administrative Expenses |
741,205 |
||||
|
Total Ordinary Expenses |
1,521,044 |
||||
|
|
Net Assets Variation of the Year |
▲954,767 |
|||
|
Net Asset brought forward |
6,390,045 |
||||
|
Net Asset carried forward |
5,435,278 |
||||
References
[1] Yudai Suzuki, So Kariyama, Yuta Ashihara, and Hiroshi Yamakawa: “Estimating Cortical Hierarchy in the Human Cerebral Cortex Using Deep Graph Matching Consensus.” In Communications in Computer and Information Science, Springer Nature Singapore, pp. 188–200, 2025. https://doi.org/10.1007/978-981-95-4100-3_14 .
[2] Yudai Suzuki, So Kariyama, Yuta Ashihara, and Hiroshi Yamakawa: “Data for Brain Reference Architecture of YS25SHCOMv0.” In Proceedings of the 3rd International Whole Brain Architecture Workshop (WBA-WS-003-02), 2025. https://drive.google.com/drive/folders/1brAW4iz7uhSoLyaOoJz_TVoUxY_nDI6U .
[3] Yohei Maruyama, Yoshimasa Tawatsuji, and Hiroshi Yamakawa: “Capability Constraints Reduce Implementation Search in Neural Circuits for Efficient Brain-Inspired Software Design.” In Proceedings of the 3rd International Whole Brain Architecture Workshop (WBA-WS-003-01), 2025. https://drive.google.com/drive/folders/1Ypsl0n4Zl1fSh0u-niNqeMBAtpvBoY_r .
[4] Naohiro Yamauchi, Yoshimasa Tawatsuji, Yudai Suzuki, Kenji Doya, and Hiroshi Yamakawa: “Data for Brain Reference Architecture of NY24CanonicalCorticalMicrocircuitsInference.” In Proceedings of the 2nd International Whole Brain Architecture Workshop (WBA-WS-002), 2025. https://drive.google.com/drive/folders/1aqIVmRxfCTDHbXcIUnNjK2RM2Gos1N2T .
[5] Naohiro Yamauchi, Yoshimasa Tawatsuji, Yudai Suzuki, Kenji Doya, and Hiroshi Yamakawa: “Data for Brain Reference Architecture of NY24CanonicalCorticalMicrocircuitsDecisionMaking.” In Proceedings of the 2nd International Whole Brain Architecture Workshop (WBA-WS-002), 2025. https://drive.google.com/drive/folders/1aRRKwsqHiis9Do5ZUvYIaofYGGYkmwVf
[6] Naohiro Yamauchi, Yoshimasa Tawatsuji, Yudai Suzuki, Hiroshi Yamakawa, and Kenji Doya: “A Computational Model of Canonical Cortical Microcircuits for Dynamic Bayesian Inference and Control as Inference.” Neuroscience Research, vol. 222, pp. 105002, 2026. https://doi.org/10.1016/j.neures.2025.105002 .
[7] Akira Taniguchi, Atsushi Fujii, Takeshi Nakashima, Tatsuya Miyamoto, Yoshimasa Tawatsuji, and Hiroshi Yamakawa: “Data for Brain Reference Architecture of AF24Hippocampus-Amygdala.” In Proceedings of the 2nd International Whole Brain Architecture Workshop (WBA-WS-002), 2025. https://drive.google.com/drive/folders/1562wm1E9JByJnOpT192bYNiF5e1CbkJO .
[8] Yoshimasa Tawatsuji: “Data for Brain Reference Architecture of YT24ITcAmy.” In Proceedings of the 2nd International Whole Brain Architecture Workshop (WBA-WS-002), 2025. https://drive.google.com/drive/folders/1ZoJn-k_oeLUBe_-fz5jSFED9kYzpq5LT .
[9] Tatsuya Miyamoto, Yoshimasa Tawatsuji, and Hiroshi Yamakawa: “Data for Brain Reference Architecture of TM25CILsSpiralingHypothesis.” In Proceedings of the 3rd International Whole Brain Architecture Workshop (WBA-WS-003-05), 2025. https://drive.google.com/drive/folders/1GsJEhtlP6uVX688b_cu_lbH5V7qAEtYe .
[10] Misaki Inoue and Yoshimasa Tawatsuji: “Data for Brain Reference Architecture of MI25Imagination.” In Proceedings of the 3rd International Whole Brain Architecture Workshop (WBA-WS-003-03), 2025. https://drive.google.com/drive/folders/1Tl2rE_h0S5O5teLIaozADmBWRo9B8ivE .
[11] Seii Ko, Yoshimasa Tawatsuji, and Tatsunori Matsui: “Data for Brain Reference Architecture of SK25LHbDepressionModel.” In Proceedings of the 3rd International Whole Brain Architecture Workshop (WBA-WS-003-04), 2025. https://drive.google.com/drive/folders/1IVt_bORkrf6qaTld5hfuF1Z9ubyjifNA .
[12] Takeshi Nakashima, Akira Taniguchi, Tadahiro Taniguchi, and Hiroshi Yamakawa: “Interpretable Anomaly Detection in a Hippocampal Formation-Inspired Spatial Cognition Model.” In Lecture Notes in Computer Science, Springer Nature Singapore, pp. 509–522, 2025. https://doi.org/10.1007/978-981-95-4384-7_35 .
[13] Hiroshi Yamakawa and Yoshimasa Tawatsuji: “The Enduring Value of Human Brain-morphic AGI in the Era of Universal Intelligence — Experiential Understanding and Partial Knowability —.” Journal of the Japanese Society for Artificial Intelligence, vol. 41, no. 2, pp. 149–156, 2026 (in Japanese). https://doi.org/10.11517/jjsai.41.2_149 .
Japanese