Reproduce “DocCGen: Document-based Controlled Code Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sameer Pimparkhede, Mehant Kammakomati, Srikanth Tamilselvam, Prince Kumar, Ashok Pon Kumar, Pushpak Bhattacharyya
DocCGen generates Bash commands and Ansible YAML by retrieving relevant library documentation and using its templates and schemas to constrain generation token by token. This makes documented syntax and field dependencies active constraints rather than context alone. In an out-of-distribution Ansible-YAML test, StarCoder2 3B achieved 58.82% Schema Correct, versus 7.47% with retrieval alone; this measures schema compliance, not runtime correctness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.