1 projects found matching your search criteria :
Alignment Of Large Language Models Via Debate And Reinforcement Learning
Call: OPUS 26 , Panel: ST6
Principal investigator: dr Łukasz Kuciński
IDEAS NCBR sp. z o.o.