Reproduce “BeSpec: Behavior-Level Specification Alignment for 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
Qinghua Xu, Guancheng Wang, Boxi Yu, Lionel Briand
BeSpec uses behavioral specifications to detect mismatches between task interpretations and programs. It converts a task into structured obligations, generates probes and candidate programs, then compares observed behavior with predicted properties to guide specification clarification. Across six evaluation settings, Qwen3-Coder-30B reaches average Pass@1 of 44.0%, versus Specine’s 35.6%, a 25.3% relative improvement; this shows benchmark gains, not verified access to task intent.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.