DIVE：在代理任务合成中扩展多样性以实现通用工具使用
DIVE: Scaling Diversity in Agentic Task Synthesis for Generalizable Tool Use
AI总结：
DIVE通过反转合成顺序，先执行多样化的现实工具并反向推导任务，从而提升代理任务合成的多样性，使模型在不同任务和工具集下的泛化能力显著增强。
摘要：
近期的研究为经过训练的工具使用LLM合成代理任务，但任务和工具集变化下的鲁棒泛化仍是一个开放性挑战。我们追溯这种脆弱性到合成任务中缺乏足够的多样性。扩大多样性是困难的，因为训练需要任务保持可执行和可验证，而泛化要求覆盖多种工具类型、工具集组合和异质工具使用模式。我们提出了DIVE，一种证据驱动的食谱，通过反转合成顺序，首先执行多样化的现实工具，然后反向推导由结果轨迹严格包含的任务，从而通过构建提供基础。DIVE在两个可控轴上扩展结构多样性，即工具集覆盖和每任务工具集多样性，并且证据收集-任务推导循环进一步在五个领域中的373种工具上诱导丰富的多步骤工具使用模式。在DIVE数据上训练Qwen3-8B（48k SFT + 3.2k RL）在9个OOD基准上平均提高22分，并优于最强的8B基线68分。值得注意的是，受控扩展分析显示，多样性扩展在OOD泛化中始终优于数量扩展，即使数据量少4倍。
英文摘要：
Recent work synthesizes agentic tasks for post-training tool-using LLMs, yet robust generalization under shifts in tasks and toolsets remains an open challenge. We trace this brittleness to insufficient diversity in synthesized tasks. Scaling diversity is difficult because training requires tasks to remain executable and verifiable, while generalization demands coverage of diverse tool types, toolset combinations, and heterogeneous tool-use patterns. We propose DIVE, an evidence-driven recipe that inverts synthesis order, executing diverse, real-world tools first and reverse-deriving tasks strictly entailed by the resulting traces, thereby providing grounding by construction. DIVE scales structural diversity along two controllable axes, tool-pool coverage and per-task toolset variety, and an Evidence Collection--Task Derivation loop further induces rich multi-step tool-use patterns across 373 tools in five domains. Training Qwen3-8B on DIVE data (48k SFT + 3.2k RL) improves by +22 average points across 9 OOD benchmarks and outperforms the strongest 8B baseline by +68. Remarkably, controlled scaling analysis reveals that diversity scaling consistently outperforms quantity scaling for OOD generalization, even with 4x less data.
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。
