Search papers, labs, and topics across Lattice.
This paper introduces Adaptive Placeholder Completion (APC), a framework that allows code completion models to strategically insert placeholders in code suggestions when uncertainty is high, rather than forcing a complete, potentially incorrect, generation. They formulate code completion as a cost-minimization problem, proving that placeholder insertion is more cost-effective than error correction above a critical entropy threshold. Experiments on models ranging from 1.5B to 14B parameters show that APC reduces editing costs by 19-50% without sacrificing standard hard completion performance.
LLMs can cut code editing costs by up to 50% simply by knowing when *not* to guess.
While Large Language Models (LLMs) have demonstrated exceptional proficiency in code completion, they typically adhere to a Hard Completion (HC) paradigm, compelling the generation of fully concrete code even amidst insufficient context. Our analysis of 3 million real-world interactions exposes the limitations of this strategy: 61% of the generated suggestions were either edited after acceptance or rejected despite exhibiting over 80% similarity to the user's subsequent code, suggesting that models frequently make erroneous predictions at specific token positions. Motivated by this observation, we propose Adaptive Placeholder Completion (APC), a collaborative framework that extends HC by strategically outputting explicit placeholders at high-entropy positions, allowing users to fill directly via IDE navigation. Theoretically, we formulate code completion as a cost-minimization problem under uncertainty. Premised on the observation that filling placeholders incurs lower cost than correcting errors, we prove the existence of a critical entropy threshold above which APC achieves strictly lower expected cost than HC. We instantiate this framework by constructing training data from filtered real-world edit logs and design a cost-based reward function for reinforcement learning. Extensive evaluations across 1.5B--14B parameter models demonstrate that APC reduces expected editing costs from 19% to 50% while preserving standard HC performance. Our work provides both a theoretical foundation and a practical training framework for uncertainty-aware code completion, demonstrating that adaptive abstention can be learned end-to-end without sacrificing conventional completion quality.
