Regression Language Models for Code
This addresses the challenge of code-to-metric regression for developers and researchers, offering a unified approach that is incremental over prior feature engineering methods.
The paper tackles the problem of predicting numeric outcomes of code executions, such as memory footprint and latency, by introducing a Regression Language Model (RLM) that achieves > 0.9 Spearman-rank on competitive programming submissions and > 0.5 average Spearman-rank across 17 languages.
We study code-to-metric regression: predicting numeric outcomes of code executions, a challenging task due to the open-ended nature of programming languages. While prior methods have resorted to heavy and domain-specific feature engineering, we show that a single unified Regression Language Model (RLM) can simultaneously predict directly from text, (i) the memory footprint of code across multiple high-level languages such as Python and C++, (ii) the latency of Triton GPU kernels, and (iii) the accuracy and speed of trained neural networks represented in ONNX. In particular, a relatively small 300M parameter RLM initialized from T5Gemma, obtains > 0.9 Spearman-rank on competitive programming submissions from APPS, and a single unified model achieves > 0.5 average Spearman-rank across 17 separate languages from CodeNet. Furthermore, the RLM can obtain the highest average Kendall-Tau of 0.46 on five classic NAS design spaces previously dominated by graph neural networks, and simultaneously predict architecture latencies on numerous hardware platforms.