SAO Outperforms GRPO on Coding and Reasoning Benchmarks
ivan_bezdomny · x · 2026-07-15
This repost points out that current improvements primarily stem from Direct Double-Sided Importance Sampling (DIS), rather than vanilla GRPO.
Key takeaways:
- Vanilla GRPO is outdated, but can still linger for a while through various tweaks.
- The new method, SAO (Single-rollout Asynchronous Optimization), appears to have a higher ceiling.
- The paper claims that SAO can stably train for 1000 steps and consistently outperforms GRPO and its variants on benchmarks like SWE-Bench Verified, BeyondAIME, and IMOAnswerBench.
This represents a training method advancement tailored for agentic coding and reasoning tasks.
Related event: SAO Outperforms GRPO in Coding and Reasoning Benchmarks(3 posts)→
More from coding & agent
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27
- BELL: Open-Source Stack Combines Browserbase, Eve, iMessage and Stripe Link for Personal Agents — jeff_weinstein · 2026-10-27
- Legora runs 2,000 parallel agent evals on Modal Sandboxes for legal AI — blelbach · 2026-10-27
- Two engineers rebuilt a FedEx supplier's billing core in 3.5 months vs an estimated 7-8 — alex_verem · 2026-10-27
- GPT-6.1 Terra scores 48 on AA Coding Agent Index at one-third the cost of GPT-5.6 Terra — ArtificialAnlys · 2026-10-27
- Shift-1 router cuts coding agent spend up to 55% while keeping SWE-bench accuracy — ypatil125 · 2026-10-27