TermiGen: 为终端智能体提供高保真的环境和鲁棒轨迹合成
TermiGen: High-Fidelity Environment and Robust Trajectory Synthesis for Terminal Agents
AI总结：
TermiGen通过生成高保真的环境和鲁棒轨迹，提升终端智能体在复杂任务中的表现，达到31.3%的通过率，超越现有基线和专有模型。
摘要：
执行复杂的终端任务仍然是开放权重大语言模型面临的重要挑战，受限于两个根本性限制。首先，高保真的可执行训练环境稀缺：从现实世界仓库中合成的环境不够多样且可扩展，而由大语言模型合成的轨迹则存在幻觉问题。其次，标准的指令微调使用专家轨迹，这些轨迹很少表现出较小模型常见的简单错误。这导致了分布不匹配，使学生模型无法从自身运行时的失败中恢复。为弥合这些差距，我们引入TermiGen，一种端到端的管道，用于合成可验证的环境和坚韧的专家轨迹。TermiGen首先通过迭代的多智能体优化循环生成功能有效的任务和Docker容器。随后，我们采用生成器-批评者协议，在轨迹收集过程中主动注入错误，合成富含错误纠正循环的数据。在TermiGen生成的数据集上微调后，我们的TermiGen-Qwen2.5-Coder-32B在TerminalBench上实现了31.3%的通过率。这建立了新的开放权重的最先进水平，优于现有基线，并显著超越了像o4-mini这样的强大专有模型。数据集可在https://github.com/ucsb-mlsec/terminal-bench-env上获取。
英文摘要：
Executing complex terminal tasks remains a significant challenge for open-weight LLMs, constrained by two fundamental limitations. First, high-fidelity, executable training environments are scarce: environments synthesized from real-world repositories are not diverse and scalable, while trajectories synthesized by LLMs suffer from hallucinations. Second, standard instruction tuning uses expert trajectories that rarely exhibit simple mistakes common to smaller models. This creates a distributional mismatch, leaving student models ill-equipped to recover from their own runtime failures. To bridge these gaps, we introduce TermiGen, an end-to-end pipeline for synthesizing verifiable environments and resilient expert trajectories. Termi-Gen first generates functionally valid tasks and Docker containers via an iterative multi-agent refinement loop. Subsequently, we employ a Generator-Critic protocol that actively injects errors during trajectory collection, synthesizing data rich in error-correction cycles. Fine-tuned on this TermiGen-generated dataset, our TermiGen-Qwen2.5-Coder-32B achieves a 31.3% pass rate on TerminalBench. This establishes a new open-weights state-of-the-art, outperforming existing baselines and notably surpassing capable proprietary models such as o4-mini. Dataset is avaiable at https://github.com/ucsb-mlsec/terminal-bench-env.
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。
