Author name: Hengyu Shi, Tianyang Han, Peizhe Wang, Zhiling Wang, Xu Yang, Junhao Su

arXiv:2605.04913v2 Announce Type: replace
Abstract: LLM post-training typically propagates task gradients through the full depth of the model. Although this end-to-end structure is simple and general, it couples task adaptation to full-depth activatio…

arXiv:2605.04913v1 Announce Type: cross
Abstract: LLM post-training typically propagates task gradients through the full depth of the model. Although this end-to-end structure is simple and general, it couples task adaptation to full-depth activation …