Generative Compilation: On-the-Fly Compiler Feedback as AI Generates Code

2026-07-15

AI Summary

This work addresses the challenge that current large language models often produce uncompilable code in strongly statically typed languages like Rust due to syntactic or semantic errors, with conventional compiler feedback arriving too late to influence the generation process. The authors propose a generative compilation approach featuring a lightweight, context-preserving “sealor” mechanism that incrementally completes partial programs into compilable forms, enabling immediate diagnostic feedback on incomplete code for the first time. Requiring no white-box access to the model and thus compatible with black-box large language models, the method integrates syntax-guided completion, constrained decoding, and the Rust compiler. Evaluated on repository-scale tasks, it substantially reduces uncompilable outputs, improves functional correctness, and enables early, precise localization of diverse error sources, effectively mitigating error cascades.

Abstract

Languages with rich static semantics, such as Rust, provide stronger guarantees for AI-generated code, but their strictness makes generation more difficult. Off-the-shelf compilers can provide useful feedback post-generation, but does not guide intermediate generation steps, such as those during autoregressive LLM decoding. Constrained decoding intervenes earlier by rejecting invalid tokens during sampling, but requires white-box model access and costly reimplementation for semantic constraints.We introduce generative compilation, the first approach to obtaining compiler feedback on partial programs during generation. The core technical device is a sealor: a lightweight, mostly syntax-guided transformation that converts partial programs into complete ones that standard compilers can diagnose. It is designed such that possible-to-complete partial programs are never rejected, while preserving enough code context to catch genuine dead ends early. We construct such a sealor on a core Rust-like calculus and prove that it satisfies these properties, all mechanized in Lean. We extend it to the first partial-program checker for real Rust. We evaluate our method on challenging repository-level Rust coding tasks, across both frontier black-box and open-weight models. We show that generative compilation reduces non-compiling outputs and improves functional correctness, relative to standard post-generation feedback. It does so by detecting a broad range of errors close to their source and early during generation, thereby reducing errors cascades and enabling focused diagnostics. More broadly, generative compilation is a step toward making compilers a first-class citizen of AI-assisted programming active during generation, rather than a separate post-generation check.

Problem

• generative compilation
• compiler feedback
• partial programs
• AI code generation
• static semantics

Innovation

• generative compilation
• sealor
• partial program checking
• compiler feedback
• constrained decoding