Meta Paper: Optimizing Code Execution Speed via RL Achieves Up to 200% Improvement
facebook · hf · 2026-07-30
Meta released research on Reinforcement Learning for Code Optimization. While using RL with hidden test cases for code correctness is established, extending it to execution speed is challenging: timing-based rewards are easily overwhelmed by measurement noise, reward sparsity, or GRPO instability.
To solve this, the researchers proposed a three-stage approach:
- Building DMC-Optim with large optimization tests and a calibrated sandbox.
- Composing correctness and speed in the RL environment, using an offline simulator to predict promising configurations.
- Adapting GRPO and evaluation for the sparser, noisier timed-execution setting.
On DMC-Optim, optimization-aware configurations improved strict top-50% pass@1 from 18.0% to 31.3% for Qwen 2.5 7B, and from 30.7% to 50.4% for CWM 32B. Under stricter criteria, improvements reach up to 200% while preserving pure-correctness scores.
Related event: Meta Research Breaks Pareto Frontier in Code Efficiency via RL(3 posts)→
More from Research
- HexRunner Robot Hits 33mph Using Pure Mechanical Design Over Sensors — lukas_m_ziegler · 2026-07-30
- NeurIPS Reviewer Ghosting: How to Handle Ignored Rebuttals — grumpket · 2026-07-30
- AIPOCH Open-Sources Library of 550+ Medical Research Agent Skills — tom_doerr · 2026-07-30
- ICSE'26 Paper Proposes New Paradigm: In-vivo Fuzzing Without Test Harnesses — moarbugs · 2026-07-30
- LMSYS Debuts Miles: Blackwell-Native 8-bit and 4-bit RL Recipes — BanghuaZ · 2026-07-30
- NeurIPS 2026 Workshop Tackles Continual Learning in Deployed AI Agents — DanielKhashabi · 2026-07-30