Mingwei Liu

Achieve near-perfect compilation rates in automated C-to-Rust translation by combining build-aware skeleton graphs with retrieval-augmented generation from historical migration data, slashing repair overhead by 60%.

LLMs can now add features to entire codebases with 36% higher success by generating multiple design options and validating their architectural impact.

LLMs struggle with new code frameworks because they lack training data capturing codebase relationships, but this work shows how to synthesize reasoning-aware data from code graphs to dramatically improve performance.