研究提出 monofact 比率决定 LLM 科学发现能力
Tracing the Scientific Discovery Capacity of LLMs to Their Training Data
训练数据里只见过一次的事实，居然直接决定模型能不能做出科学发现，两个开源模型上验证了，做数据配比的值得看。
一项 arXiv 论文把 LLM 的科学发现能力归因于训练数据中的 monofact（训练时只出现一次的事实）比率。在材料科学与调控基因组学两个领域，研究者直接调整训练集中的 monofact 比率，发现更多 monofact 在两个开源权重模型家族、不同采样策略和推理预算下都带来更多科学发现。实验还表明该效应无法仅用训练集多样性解释，且与领域内数据搭配时可跨科学领域迁移。
Tracing the Scientific Discovery Capacity of LLMs to Their Training Data
Large language models (LLMs) show growing promise for scientific discovery, yet basic questions remain about what enables reliable discovery. How can we trace the scientific discovery capacity of LLMs to their training data? We study this question as one of missing probability mass and develop a theoretical account in which monofacts-the facts observed once during training-set the discovery capacity. In materials science and regulatory genomics, we directly manipulate monofact rates in training sets and find that more monofacts consistently yield more discoveries across two open-weight model families, sampling strategies, and reasoning budgets. Further experiments show the relationship cannot be explained by training-set diversity alone and can transfer across scientific domains when paired with in-domain data. Together, our results identify monofact rate in training data as a key factor governing discoveries in language models.