Beyond Fixed Groups: Adaptive Rollouts for Coding-Agent Reinforcement Learning

Long-horizon coding-agent rollouts have highly variable latency: an agent may search a repository, edit files, and run several test suites before receiving one verifiable reward. Fixed-size GRPO allocates the same number of rollouts to every query, while round-based collection delays resampling and fresh work behind stragglers. We present ARCA, an event-driven realization of Adaptive Rollouts for Coding Agents. It processes each query when its current sampling stage completes, closes mixed groups, resamples all-wrong groups within a bounded budget, and refills fresh queries. The policy stays fixed during collection; the learner updates after collection and draining. In single-run experiments with Qwen3.6-27B on SWE-bench Verified, the $(8,6,4)$ schedule at fresh-query width $b=16$ records $2.44$ rollout steps/h versus $0.87$ for native $N=16$ GRPO ($2.80\times$). Observed peak pass@1 on the 51-problem evaluation slice is $0.608$ versus $0.510$, a gain of $9.8$ percentage points. At $b=8$, $(8,6,4)$ reaches $0.627$ peak pass@1, while $(8,4,4)$ records the highest throughput, $3.04$ steps/h. These throughput comparisons use different mixed-group targets and do not measure total compute savings.