Researchers test AI code agents on multi-repository real-world tasks

Original: Can current code agents survive beyond single-repo bug fixing?

Source: arxiv.org ↗