Users expressed gratitude for mentorship and collaboration with researchers like Taylor Killian on studies showing how SFT and RL advance AI reasoning generalization and compositional capabilities.
Based on 2 visible X reactions from 1 accounts; directional sample.
Ask a question below.
Published answers will appear here.
9/9 Fortunate to work with & learn from Xin Liu, Guangyi Chen, @martinmamql, @XiangchenSong, Yuekai Sun, @Yurochkin_M, @tw_killian, @rsalakhu, @kunkzhang, @EricXing, and @waterluffy. Happy to chat more!
@tw_killian Thank you Taylor for your kind words and incredible mentorship!
This was the primary hypothesis we sought to elucidate in this project. Pre-training creates the general distribution of possible concepts, SFT demonstrates how logical concepts can be ordered, RL discovers novel orderings when extrapolating from new context/tasks/questions. https://twitter.com/lingjingkong/status/2075400262924001683
This is some fantastic work that @LingjingKong has at ICML this week! I hope that anyone lucky enough to be in Seoul at the conference gets the chance to chat with him about what we've learned! https://twitter.com/IFM_MBZUAI/status/2075270056481615963
Users expressed gratitude for mentorship and collaboration with researchers like Taylor Killian on studies showing how SFT and RL advance AI reasoning generalization and compositional capabilities.
Based on 2 visible X reactions from 1 accounts; directional sample.
Ask a question below.
Published answers will appear here.