The results suggest that random weight perturbations may be ineffective at uncovering task experts around pretrained models, and that task adaptation requires more sophistication than random search.
On-policy distillation (OPD) is emerging as an important paradigm for language-model post-training, combining dense teacher supervision with training on student-generated trajectories. This creates a feedback loop where each update changes the student and thereby the training distribution that generates subsequent upda...
Guan-Zhe Hong, S. Motwani, Liang-Qiang Chen et al.· 0 citations
This work derives a token- and layer-wise decomposition of how learning from one token changes another prediction, and connects data attribution, forgetting, and plasticity loss as distinct regimes of the same evolving learning dynamics.
Yi Ren, Wen-Long Deng, Guan-Zhe Hong et al.· arXiv.org· 1 citation
It is shown that preference alignment preserves the human response distribution only under a restrictive condition, and no consistent evidence that real human preferences satisfy it, and human-likeness is established as an explicit dimension of alignment rather than something assumed to follow from preference alignment...
Su-Qin Yuan, Runqi Lin, Mu-Yang Li et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.