Recent advancements in large language model alignment leverage token-level supervisions to perform fine-grained preference optimization. However, existing token-level alignment methods either optimize ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback