Search papers, labs, and topics across Lattice.
This paper introduces RLPF, a novel approach to training code generation models that incorporates performance feedback to prioritize not just correctness but also runtime efficiency. By employing a staged reward system that ranks execution outcomes, RLPF effectively guides code agents to produce faster implementations while still ensuring correctness. Fine-tuning the Qwen3-32B model with RLPF on PerfCodeBench significantly improves the rate of correct and runnable solutions from 11.1% to 54.6%, while also enhancing relative efficiency from 8.1% to 38.6%, demonstrating the potential for more competitive code generation systems.
RLPF transforms how code generation models are trained by prioritizing runtime efficiency alongside correctness, leading to a dramatic increase in both runnable solutions and execution speed.
Code models are increasingly trained with execution feedback, but most training signals still stop at correctness. This leaves an important gap for systems code: two programs can pass the same tests while differing greatly in runtime. We study how to train code agents to prefer faster correct implementations, rather than treating efficiency only as an evaluation metric. The key difficulty is that runtime is a fragile reward. It is meaningful only after a program is correct, varies across tasks, and gives little guidance when most sampled programs fail to compile or run. We propose \textbf{RLPF}, reinforcement learning from performance feedback, which turns execution outcomes into a staged reward. Failed programs are ordered by execution progress, while correct programs are ranked by their relative improvement from the baseline toward the expert reference. This gives useful feedback before correctness and performance-sensitive feedback after correctness. Fine-tuning Qwen3-32B with RLPF on PerfCodeBench raises correct-and-runnable solutions from $11.1\%$ to $54.6\%$ and improves relative efficiency from $8.1\%$ to $38.6\%$. The trained model becomes competitive with stronger open-weight systems, and its optimization behavior transfers modestly to EffiBench-X. Additional studies show that model-generated references provide useful but weaker supervision, and that the full composite reward is more reliable than correctness-only or runtime-only baselines. These results suggest that code agents can be trained not only to pass tests, but also to optimize the programs they write.
