AIM-Harvard/

proof-of-time

This dataset contains benchmarks for evaluating LLM agents on academic paper analysis tasks that require understanding research trends, citations, and future directions. All evaluation data uses post-training-cutoff (2025) papers to avoid data contamination.

Paper: Proof of Time: A Benchmark for Evaluating Scientific Idea Judgments Repository:… See the full description on the dataset page: https://huggingface.co/datasets/AIM-Harvard/proof-of-time.