Fetching the paper…

GDPO: Learning to Directly Align Language Models with Diversity Using GFlowNets · Around