#kaist

AI

KAIST and Microsoft Research Asia Develop AI Error Correction via Brain Signals

Researchers from KAIST and Microsoft Research Asia have created a brain-computer interface that allows AI to adjust its actions by interpreting human brain signals during errors.

October 6, 2026, 11:58 PM·iXBT.com
AI

AI Reasoning Steps Leave Detectable Internal Patterns

A joint study by KAIST and Naver AI Lab demonstrates that AI reasoning steps correspond to distinct internal numerical patterns, most notably within the middle layers of language models.

The Decoder
That is everything