This year’s edition of the Workshop on Cognitive Architectures will be held in conjunction with the 52nd International Symposium on Computer Architecture (ISCA 2025). The full-day workshop will be held on Sunday, June 22, 2025, in Tokyo, Japan, and is organized by IBM Research.
The CogArch workshop emphasizes the challenges associated with the implementation of generative AI and the integration of chiplets as a means to fully realize its potential. As generative AI models continue to expand in size and complexity, the resulting computational demands impact the entire software-hardware ecosystem. This creates a variety of new challenges that necessitate unconventional strategies to maintain scalability in both upward and outward directions. With Large Language Model (LLM) parameter sizes approaching several billions, chiplet-based architectures represent a promising technological advancement that could enable a cost-effective and energy-efficient solution for processing such models, thus potentially transforming the future landscape of cognitive systems.
The CogArch workshop solicits formative ideas and new product offerings in the general space of AI systems that covers all the design aspects of cognitive systems, with particular focus this year on the adoption of chiplets as a promising way to support large-scale generative AI.
Topics of interest include (but are not limited to):
- 2.5D/3D chiplet architectures, along with wafer scaling and various heterogeneous integration methods, including optical heterogeneous integration, to create scalable frameworks for generative AI models.
- Development of software and compiler frameworks for large-scale deployment of generative AI models.
- Hardware-software co-design for commercially deployed AI hardware acceleration frameworks.
- Accelerators and micro-architectural support for LLMs.
- Reliability and safety considerations, and security against adversarial attacks in cognitive architectures.
- Techniques for improving energy efficiency of AI applications, and battery life extension and endurance in mobile AI architectures.
- AI/ML for fast system modeling and AI/ML as design methodology.
- Privacy-preserving inference on AI models.
- Prototype demonstrations in specific application domains: e.g., natural language processing and speech, protein folding, drug discovery, computer vision, code generation, music making, as well as applications of interest to defense and homeland security.
→ Accepted works will have the chance to be published in one of the leading journals in computer architecture.
Important dates:
- Paper submission deadline: May 2, 2025
- Notification of acceptance: May 16, 2025
- Workshop date: June 22, 2025
Organizers:
- Pradip Bose (IBM Research)
- Alper Buyuktosunoglu (IBM Research)
- Eri Ogawa (IBM Research – Tokyo)
- Mori Ohara (IBM Research – Tokyo)
- Karthik Swaminathan (IBM Research)
- Augusto Vega (IBM Research)