ProgramDistill: From Interactive Web Apps to Verifiable Reference-Guided SWE Tasks
Coding agents are typically evaluated with desired behavior specified through issues or instructions. In practical web development, however, agents may need to infer behavior from working software and implement it in an incomplete application. We introduce ProgramDistill, a benchmark evaluating coding agents on feature...
Jeonghye Kim, Minseon Kim, Young Jin Kim et al.
· 0 citations