ParamΔ for Direct Mixing: Post-Train Large Language Model At Zero Cost.

Sheng Cao, Mingrui Wu, Karthik Prasad, Yuandong Tian, Zechun Liu

Browse the full ICLR paper archive.