Cross-Benchmark Transfer from RL on Agentic Coding Tasks
Sushant Mehta, Logan Ritchie and Edwin Chen at Surge AI post-train Kimi K2.7 Code with RL alone on 1,700 expert-built coding tasks and measure gains on six external benchmarks and on harnesses never used in training.
Ask this paper
Training data and reward. 1,000 repository tasks graded by hidden fail-to-pass tests plus pass-to-pass tests of existing behavior, and 700 terminal tasks with expert-written hidden verifiers. Reward is the fraction of target checks passed and drops to zero if any existing-behavior test breaks.
Recipe. One epoch of GSPO on a rank-32 LoRA adapter over a 1T-parameter (32B active) MoE model.
Results. pass@1 rises on every external benchmark: SWE-Bench Pro 60.1 to 64.8, DeepSWE 31.0 to 43.4, Terminal-Bench 2.1 67.4 to 82.0, Terminal-Bench 3 1.4 to 12.1, Terminal-Bench 4 0.0 to 7.6, SWE-Marathon 5.0 to 25.0.
Contamination and harness checks. The pooled gain is significant (p < 0.001) and stays significant on the three sets released after the training data was collected (p = 0.004). The model also improves under two harnesses absent from training.
What changed. Median trajectories are 24 to 35% shorter. Most base-model failures were near-misses, and on newly solved tasks the trained model avoids dropped requirements, tests that only cover the implemented path, broken existing behavior and unchecked assumptions.
Abstract
Coding agents often fail in the last mile: they build most of a feature but drop a requirement, test only the cases their implementation already handles, break behavior that was supposed to stay intact, or validate against an unchecked assumption. We ask whether reinforcement learning (RL) on expert-built agentic coding tasks closes this gap, and whether what the agent learns transfers beyond the training distribution. We post-train Kimi K2.7 Code, a 1T-parameter (32B active) open-weight mixture-of-experts model, with RL alone on 1,700 tasks: 1,000 repository tasks graded by hidden fail-to-pass tests and by pass-to-pass tests of existing behavior, and 700 terminal tasks graded by expert-written hidden verifiers. The reward is the fraction of target checks passed and drops to zero if any pass-to-pass test fails. One epoch of GSPO on a rank-32 LoRA adapter improves pass@1 on each of the six external benchmarks we evaluated, across three agent harnesses: SWE-Bench Pro (60.1 to 64.8), DeepSWE (31.0 to 43.4), Terminal-Bench 2.1 (67.4 to 82.0), Terminal-Bench 3 (1.4 to 12.1), Terminal-Bench 4 (0.0 to 7.6), and SWE-Marathon (5.0 to 25.0). Pooled over the five independent task sets (Terminal-Bench 4 revises Terminal-Bench 3), the improvement is significant (p < 0.001), and it remains significant on the three sets released after the training data was collected (p = 0.004); the model also improves under both harnesses never used in training. Median trajectories on DeepSWE and Terminal-Bench 3 are 24-35% shorter in agent steps. The base model's failed DeepSWE runs are mostly near-misses, and on the tasks the trained model newly solves, paired trajectories show it avoiding each of the four failure modes above.