arXiv cs.CL· Teng Pan, Zixuan Wang, Yuchen Yan, Qifei Wu, Yao Chen, Yingqi Qu, Ruiqing Zhang, Yongliang Shen·· 11 小时前AI 评分38
Agent-R²：用真实来源数据合成扩展智能体，4B 模型超越 Qwen3-235B-A22B-Thinking-2507
Agent-R$^2$: Scaling Agents for Realistic Productivity Work via Real-Source Data Synthesis
AI 导读
Agent-R² 是一个面向真实生产力工作的真实来源数据合成框架，从 1 万个网络语料候选中蒸馏出 4k 个任务实例，配以真实多格式附件和交付物级评分标准。
正文 · 原文
Abstract:Large language models have made rapid progress in agentic reasoning and tool use, raising expectations for agents that can take over realistic productivity work from humans. However, current agents still struggle to finish such tasks, where success requires following long-horizon professional workflows, understanding heterogeneous file contexts, and producing usable deliverables under fine-grained requirements. A key bottleneck is the lack of scalable post-training data that reflects real work intent, authentic file environments, and deliverable-level evaluation. To close this gap, we introduce Agent-R$^2$, a $\textbf{r}$eal-source data synthesis framework that scales training data for $\textbf{r}$ealistic productivity work. Instead of relying on manually specified workflows or LLM-generated files, Agent-R$^2$ grounds synthesis throughout the pipeline in real occupational seeds, corpus-derived workflow evidence, and authentic web-sourced files. From 10k web-corpus candidates, Agent-R$^2$ distills 4k task instances that pair real multi-format attachments with fine-grained, deliverable-level rubrics. On this data, we train Agent-R$^2$-4B via supervised fine-tuning and rubric-based reinforcement learning, improving both end-to-end productivity performance and file-grounded reasoning. Trained on only 4k instances, Agent-R$^2$-4B surpasses Qwen3-235B-A22B-Thinking-2507 on both GDPval (51.0$\%$ vs. 48.0$\%$) and OfficeQA (14.8$\%$ vs. 13.8$\%$). Further analysis shows that Agent-R$^2$ mitigates shortcut behavior, where agents produce superficially plausible deliverables before sufficiently grounding, constructing, or verifying them, while promoting sustained exploration.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.08219 [cs.CL] |
| (or arXiv:2611.08219v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.08219 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Teng Pan [view email]
[v1]
Thu, 5 Nov 2026 16:08:05 UTC (2,732 KB)
来源：arXiv cs.CL · arxiv.org