Towards Efficient LLMs Annealing with Principled Sample Selection 事件

Towards Efficient LLMs Annealing with Principled Sample Selection arXiv:2605.31175v1 Announce Type: new Abstract: The annealing phase is a pivotal convergence stage in LLM pre-training that ultimately determines final model quality. However, effectively selecting training data during this phase remains a key challenge. Current strategies rely on empirical heuristics, such as domain filtering or context extension, which lack a principled grounding in optimization theory. In this work, we charact