🧠 New arXiv paper: “Can Language Models Rebuild Programs From Scratch?” / ProgramBench evaluates whether models can recreate real codebases from binaries and docs. arxiv.org/abs/2605.03546 #AI #LLM #ProgramBench #Research #MachineLearning #SoftwareEngineering

@DotCSV que opinas del nuevo benchmark de Meta? #ProgramBench arxiv.org/abs/2605.03546