1
MIRA: Mid-training Rubric Anchoring for Source-Aware Data Selection
提出MIRA方法,通过中训练评分锚定实现来源感知数据选择,提升大模型训练质量。
arXiv:2605.30288v1 Announce Type: new Abstract: Mid-training has become an important stage in modern LLM development, using large-scale curated mixtur…