Search papers, labs, and topics across Lattice.
This paper introduces OpenCoder, an uncertainty-aware framework for retrieval-augmented code generation that models the uncertainty in heterogeneous evidence sources like similar-code examples and repository context. By filtering and ranking evidence based on source-specific uncertainty, OpenCoder enhances the generation, verification, and repair of code, achieving a significant increase in output correctness from 56.25% to 78.13% in a 32-task evaluation. The findings reveal that the effectiveness of evidence sources is context-dependent, highlighting the importance of treating uncertainty as a control signal in code generation tasks.
Uncertainty in retrieval can be harnessed to boost code generation accuracy by over 20% when properly managed.
Repository-level code generation relies on heterogeneous evidence whose relevance, compatibility, and completeness are inherently uncertain. Similar-code examples, repository context, and project-specific APIs may provide complementary information, but can also introduce noisy, redundant, or conflicting signals. Existing retrieval-augmented approaches primarily optimize retrieval relevance without explicitly modeling how uncertainty in retrieved evidence affects downstream generation. We introduce OpenCoder, an uncertainty-aware framework that estimates source-specific uncertainty, uses it to filter and rank heterogeneous evidence, and guides generation, verification, and repair. A factorial analysis over API knowledge, repository context, and similar-code evidence reveals no universal additive source ranking; instead, significant cross-source interactions depend on the accompanying evidence and LLM backend. On an expanded 32-task RepoExec-inline evaluation, OpenCoder improves GPT selected-output correctness over Baseline RAG from 56.25\% to 78.13\%. However, it matches a verification-and-repair control, and the corresponding Gemini improvement is not statistically supported, indicating backend-dependent benefits. Target-aware API refinement also substantially improves API-set retrieval. These findings support treating uncertainty as an actionable control signal for repository-level retrieval, verification, and repair.
