Topic
scientific creativity
LLM Jaggedness Unlocks Scientific Creativity: New Benchmark Reveals Uneven AI Capabilities Can Be Harnessed for Innovation
A new arXiv paper introduces SciAidanBench, a benchmark for measuring the scientific creativity of large language models. The research finds that LLM capabilities are jagged—uneven across tasks and domains—but that this jaggedness can be harnessed through ensemble methods to produce superior scientific ideas.
SciText2Eq Study: LLMs Show Limited Accuracy in Generating Equations from Scientific Text for Enterprise AI
A new paper, SciText2Eq, evaluates large language models (LLMs) on generating mathematical equations from scientific texts. The study constructed a dataset from AI research papers and introduced a multi-faceted evaluation protocol. Results show that LLMs achieve only moderate lexical similarity and suffer from poor semantic accuracy, and that LLM-based evaluations correlate poorly with human judgments, highlighting challenges for reliable AI in technical domains.