Selected:
A new way to increase the capabilities of large language models
December 18, 2025
MIT-IBM Watson AI Lab researchers developed an expressive architecture that provides better state tracking and sequential reasoning in LLMs over long texts.