ByteDance Seed’s HarnessDev project tests if large language models can autonomously engineer robust agent harnesses, revealing a significant generalization gap.
Can LLMs Create Generalizable Agent Harnesses? ByteDance Seed’s Findings Unveiled
