Answer: Fujitsu announced a “self-evolving multi-AI agent technology” designed for teams of agents to learn continuously and safely from execution results, human feedback, policy revisions, and specification changes. Fujitsu said the agents identify reasons for success and failure, extract actionable knowledge, and take over tasks such as prompt adjustments and evaluation-criteria updates that previously required experts.
Fujitsu said it applied the approach to automate and optimize steps in building business-specific LLMs, including data selection, learning-condition tuning, evaluation, and improvement. The company said it automatically enhanced its “Takane” across manufacturing, healthcare, finance, and public administration, and achieved an average accuracy improvement of 28 points compared with pre-specialization performance.
Source: Fujitsu Global