CHART: A Harness-Rotation Curriculum for Harness-Robust Search Agents

arXiv 2609.22247 ↗First posted 5 Sep 2026cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 32 of its 34 references that are in the map. First posted on arXiv 5 Sep 2026.

Builds on

32 of 34 references.

References · 32

2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
2025Qwen3 Technical Reportcited by 8,195
2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,747
2021Multitask Prompted Training Enables Zero-Shot Task Generalizationcited by 1,462
2025Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learningcited by 1,371
2021Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivitycited by 1,122
2022Large Language Models Are Human-Level Prompt Engineerscited by 1,020
2023Large Language Models as Optimizerscited by 830
2023Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formattingcited by 778
2025GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learningcited by 449
2025R1-Searcher: Incentivizing the Search Capability in LLMs via Reinforcement Learningcited by 322
2023State of What Art? A Call for Multi-Prompt LLM Evaluationcited by 299

20 more

2026Meta-Harness: End-to-End Optimization of Model Harnessescited by 248
2024WebRL: Training LLM Web Agents via Self-Evolving Online Curriculum Reinforcement Learningcited by 243
2023PromptRobust: Towards Evaluating the Robustness of Large Language Models on Adversarial Promptscited by 218
2025ZeroSearch: Incentivize the Search Capability of LLMs without Searchingcited by 203
2023EvoPrompt: Connecting LLMs with Evolutionary Algorithms Yields Powerful Prompt Optimizerscited by 189
2023An LLM Compiler for Parallel Function Callingcited by 170
2025ToRL: Scaling Tool-Integrated RLcited by 166
2024Fact, Fetch, and Reason: A Unified Evaluation of Retrieval-Augmented Generationcited by 158
2026Kimi K3: Open Frontier Intelligencecited by 99
2025SealQA: Raising the Bar for Reasoning in Search-Augmented Language Modelscited by 65
2023Evaluating the Zero-shot Robustness of Instruction-tuned Language Modelscited by 63
2024Training Large Language Models for Reasoning through Reverse Curriculum Reinforcement Learningcited by 63
2025SwS: Self-aware Weakness-driven Problem Synthesis in Reinforcement Learning for LLM Reasoningcited by 39
2025VCRL: Variance-based Curriculum Reinforcement Learning for Large Language Modelscited by 35
2026Brittlebench: Quantifying LLM robustness via prompt sensitivitycited by 19
2024Revisiting OPRO: The Limitations of Small-Scale LLMs as Optimizerscited by 14
2026Group Distributionally Robust Optimization-Driven Reinforcement Learning for LLM Reasoningcited by 9
2025When Models Can't Follow: Testing Instruction Adherence Across 256 LLMscited by 5
2026MAS-PromptBench: When Does Prompt Optimization Improve Multi-Agent LLM Systems?cited by 4
2025Reinforcement Learning for Long-Horizon Multi-Turn Search Agentscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way