超越下一个词预测：Atlassian工作流中工具使用代理的RLVR概念验证
Beyond Next-Token Prediction: An RLVR Proof of Concept for Tool-Use Agents on Atlassian Workflows
AI总结：
针对企业SaaS工作流中LLM因目标不匹配导致的静默失败，提出在目标环境中直接应用可验证奖励强化学习（RLVR），在五个合成Jira/Confluence环境中训练Qwen模型，将平均奖励从0.35-0.92提升至0.95-1.00。
摘要：
大型语言模型被训练来预测下一个词，而不是在特定API内部行动。在利基企业SaaS工作流中——成功意味着以正确的顺序用正确的嵌套参数命中正确的端点——这种目标不匹配表现为静默失败：遗漏必填字段、幻觉工具或在单次读取后提前停止。我们询问直接应用于目标环境的可验证奖励强化学习（RLVR）是否能弥合这一差距。作为概念验证，我们构建了一套五个合成环境，以模式保真度模拟Jira REST v3和Confluence v2 API；奖励完全从工具调用轨迹计算，无需实时API、无需学习型评判者、无需人工标注。在驱动GRPO训练的相同检查器上对Qwen3-1.7B和Qwen3.5-4B进行评分，我们发现，在奖励非退化的四个场景中，RL训练的策略将平均奖励从4B基线范围0.35-0.92提升至0.95-1.00，其中Confluence页面创建场景的提升最大（0.35→1.00）。我们将此定位为面向利基企业API的成果优化小模型的初步步骤，并强调研讨会读者应权衡的两个局限性：手工制作可验证奖励无法扩展到本文报告的少数端点之外，并且我们的五个场景之一（工单转换）具有饱和奖励形状，提示的4B模型已经达到最大值。
英文摘要：
Large language models are trained to predict the next token, not to act inside a specific API. In niche enterprise SaaS workflows -- where success means hitting the right endpoint with the right nested arguments in the right order -- this objective mismatch shows up as silent failures: dropped required fields, hallucinated tools, or early stops after a single read. We ask whether Reinforcement Learning with Verifiable Rewards (RLVR), applied directly in the target environment, closes the gap. As a proof of concept we build a suite of five synthetic environments emulating the Jira REST v3 and Confluence v2 APIs at schema fidelity; rewards are computed entirely from the tool-call trace, with no live API, no learned judge, and no human label in the loop. Scoring prompted Qwen3-1.7B and Qwen3.5-4B on the same checkers that drive GRPO training, we find that on the four scenarios whose rewards are non-degenerate the RL-trained policy lifts average reward from a 4B-baseline range of 0.35--0.92 to 0.95--1.00, with the largest single gain on Confluence page creation ($0.35 \rightarrow 1.00$). We position this as a preliminary step toward outcome-optimised small models for niche enterprise APIs, and foreground two limitations a workshop reader should weigh: hand-crafting verifiable rewards does not scale beyond the handful of endpoints reported here, and one of our five scenarios (ticket-transition) has a saturating reward shape that the prompted 4B already maxes out.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
