To the future.

We built a benchmark to measure how well LLMs can extract experiments from materials science papers, and found that most pipelines task LLMs incorrectly, leading to inaccurate extractions.

Building robust automation by adding flexibility in place of complexity.