Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
University of Dayton
1
0
2
Current open code LLMs are far from human-level performance, with the best model only achieving 23.64% correctness and revealing critical failure modes that traditional evaluations miss.