CHART: A Harness-Rotation Curriculum for Harness-Robust Search Agents
Xinlu Zhang, Besnik Fetahu, Xi Chen and colleagues at Amazon show that a search agent trained with GRPO under one harness learns parallel search only for that harness, and propose CHART, a rotating harness curriculum that makes the behavior hold across prompt rewrites (NeurIPS 2026 CL4FMAgents workshop).
Ask this paper
Harness-local behavior. Qwen3-30B-A3B-Thinking trained under a fixed system prompt falls back to serial search when the prompt is rewritten, even though the task is unchanged.
Why augmentation fails. GRPO learns from the reward gap between parallel and serial rollouts of the same question; a small harness pool saturates that gap early, and a large pool spreads the signal too thinly for any harness to consolidate.
Rotation curriculum. At each periodic evaluation CHART retires ("graduates") harnesses whose behavior is learned and swaps in harnesses that are still learnable, so the reward gap stays open.
Results. From the same pool, CHART learns parallel search on every harness while static augmentation succeeds on at most half; on held-out harnesses it parallelizes 89% of turns against at most 5% for static pools.
Transfer. On a new QA task and search environment pass@1 rises 5.6 points over the best static pool, and CHART-trained agents gain more from later meta-harness search.
Abstract
Search agents are usually trained under a single harness. But once an agent is deployed in a real application, its harness is frequently updated (e.g., a rewritten system prompt) to fit production needs. This exposes a fragility of post-trained agents: because a learned behavior is entangled with its training harness, even a harness update that leaves the task unchanged can fail to elicit the behavior. We train a search agent to perform parallel search, a popular strategy for improving both search efficiency and performance. We find that training under a fixed harness makes the behavior harness-local, overfit to that harness's surface form: when the harness changes, the model falls back to serial search. An intuitive fix is harness augmentation, but simply training on more harnesses does not resolve the problem. GRPO learns from the reward gap between parallel and serial rollouts of the same question: a small harness pool saturates that gap early, while a large pool dilutes the per-harness signal too thinly for any harness to consolidate. We therefore propose Curriculum HArness Rotation Training (CHART), a rotating curriculum that lets a search agent gradually consolidate parallel search across harnesses. At each periodic evaluation, CHART "graduates" the harnesses whose expected behavior is learned and replaces them with still-learnable ones, keeping the reward gap alive throughout training. Starting from the same harness pool, CHART makes the model learn parallel search on all harnesses, whereas static augmentation succeeds on at most half of them. The behavior also carries to held-out harnesses: CHART parallelizes on 89% of held-out turns, against at most 5% for the static pools. It further transfers to a new QA task and search environment, improving pass@1 by 5.6pp over the best static pool. Finally, CHART-trained agents benefit more from meta-harness search than baselines.