LEGO-Compiler: Enhancing Neural Compilation Through Translation Composability

2025-05-26

AI Summary

Current large language models (LLMs) exhibit limited reliability in long-range, complex program translation tasks, hindering their deployment in production-grade compilers and code transformation tools. To address this, we propose LEGO, a novel neural compilation paradigm that formally establishes the composability of program translation. LEGO modularizes the compilation pipeline into verifiable, testable sub-steps and introduces an external-execution-feedback-driven self-correction mechanism. Our approach integrates program chunking and reassembly, verifiable workflow modeling, and test-driven feedback learning. Evaluated on ExeBench and AnsiBench, LEGO achieves 99.0% and 97.9% translation accuracy, respectively, while enabling near 10× scaling in compilable program size. These results significantly improve LLMs’ reliability and scalability for assembly-level program translation, advancing their practical applicability in compiler infrastructure.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Large language models (LLMs) have the potential to revolutionize how we design and implement compilers and code translation tools. However, existing LLMs struggle to handle long and complex programs. We introduce LEGO-Compiler, a novel neural compilation system that leverages LLMs to translate high-level languages into assembly code. Our approach centers on three key innovations: LEGO translation, which decomposes the input program into manageable blocks; breaking down the complex compilation process into smaller, simpler verifiable steps by organizing it as a verifiable LLM workflow by external tests; and a feedback mechanism for self-correction. Supported by formal proofs of translation composability, LEGO-Compiler demonstrates high accuracy on multiple datasets, including over 99% on ExeBench and 97.9% on industrial-grade AnsiBench. Additionally, LEGO-Compiler has also acheived near one order-of-magnitude improvement on compilable code size scalability. This work opens new avenues for applying LLMs to system-level tasks, complementing traditional compiler technologies.

Problem

• Handles long complex programs via block decomposition
• Breaks compilation into verifiable LLM workflow steps
• Improves compilable code size scalability significantly

Innovation

• Decomposes programs into manageable LEGO blocks
• Organizes compilation as verifiable LLM workflow
• Uses feedback mechanism for self-correction