scientific-forecasting

本文提出一种配对预测协议，用于衡量科学解释对实验预测的额外价值，并在 Tox21 与 OpenML 基准上测试；结论是自然解释的预测信用在测试分辨率下仍未得到确认，而 DeepSeek V4 系列模型的解释卡可降低部分次级漂移。

This paper introduces SCoR, a hierarchical framework for forecasting relations between scientific concepts, with a benchmark and model that improve research-direction discovery by predicting typed relations.

Introduces SciPaths, a benchmark for forecasting the enabling contributions required to realize a target scientific discovery, and evaluates frontier and open-weight language models, finding significant room for improvement in reasoning backward from contributions to enabling building blocks.