1 paper · 1 filter
Ran Xu, Tianci Liu, Zihan Dong +6
Standard reward models typically predict scalar scores that fail to capture the multifaceted nature of response quality in non-verifiable domains, such as creative writing or open-…