BenchmarkDeepInstruct数据集
DeepInstruct
A controlled pairwise comparison dataset for evaluating research idea quality, covering three core dimensions: novelty, significance, and feasibility. It serves as a standard benchmark for training and validating models that assess the quality of scientific research ideas.
Timeline (last 90 days)
Sep 22
The researchers introduced the DeepInstruct dataset, which provides controlled pairwise comparisons across three core dimensions: novelty, significance, and feasibility.
Unverified50%