Keio University
Keio University
03/30/2026 | Press release | Distributed by Public on 03/29/2026 20:19
Avoiding AI Blunders―Fine-Tuning-Free AI Alignment Inspired by Collision Avoidance Technology―
Avoiding AI Blunders―Fine-Tuning-Free AI Alignment Inspired by Collision Avoidance Technology―
March 6, 2026
Keio University
Associate Professor Masaki Inoue and Yuya Miyaoka (2nd-year master's student) of the Keio University Department of Applied Physics and Physico-Informatics have developed an add-on alignment technology that adjusts artificial intelligence (AI) to meet desired ethical "norms" and values.
The main feature of this technology is that it can freely adjust AI output simply by adding an external filter, requiring no fine-tuning of the foundation model itself. This enables the flexible and low-cost development of AI tailored to specific environments and applications.
This alignment technology accelerates the safe utilization of advanced AI in physical and social domains, ranging from physical systems such as robotics where safety is essential, to social infrastructure systems that require high levels of sustainability and security such as power, transportation, and communications.
The outcomes of their research were published online as a rapid release in IEEE Transactions on Control Systems Technologyon 27 March 2026.
For further information, please refer to the following PDF file.
Avoiding AI Blunders―Fine-Tuning-Free AI Alignment Inspired by Collision Avoidance Technology―
Avoiding AI Blunders―Fine-Tuning-Free AI Alignment Inspired by Collision Avoidance Technology―
Keio University published this content on March 30, 2026, and is solely responsible for the information contained herein. Distributed via Public Technologies (PUBT), unedited and unaltered, on March 30, 2026 at 02:19 UTC. If you believe the information included in the content is inaccurate or outdated and requires editing or removal, please contact us at [email protected]