Search papers, labs, and topics across Lattice.
Affiliation:
2
0
3
PhySciBench reveals that top LLMs struggle with scientific reasoning, achieving only 33.5% accuracy, while DelveAgent demonstrates a promising 7.5% improvement in performance.
