From Completion to Editing: Unlocking Context-Aware Code Infilling via Search-and-Replace Instruction Tuning
This addresses the problem of inefficient and error-prone code completion for developers by offering a more dynamic and context-aware editing approach, though it appears incremental as it builds on existing Chat LLM and agentic workflow concepts.
The paper tackles the limitations of existing code completion methods by proposing Search-and-Replace Infilling (SRI), a framework that integrates verification-and-editing into a single-pass process, enabling Chat models to surpass Base model performance with minimal data (20k samples) while maintaining low latency.
The dominant Fill-in-the-Middle (FIM) paradigm for code completion is constrained by its rigid inability to correct contextual errors and reliance on unaligned, insecure Base models. While Chat LLMs offer safety and Agentic workflows provide flexibility, they suffer from performance degradation and prohibitive latency, respectively. To resolve this dilemma, we propose Search-and-Replace Infilling (SRI), a framework that internalizes the agentic verification-and-editing mechanism into a unified, single-pass inference process. By structurally grounding edits via an explicit search phase, SRI harmonizes completion tasks with the instruction-following priors of Chat LLMs, extending the paradigm from static infilling to dynamic context-aware editing. We synthesize a high-quality dataset, SRI-200K, and fine-tune the SRI-Coder series. Extensive evaluations demonstrate that with minimal data (20k samples), SRI-Coder enables Chat models to surpass the completion performance of their Base counterparts. Crucially, unlike FIM-style tuning, SRI preserves general coding competencies and maintains inference latency comparable to standard FIM. We empower the entire Qwen3-Coder series with SRI, encouraging the developer community to leverage this framework for advanced auto-completion and assisted development.