Tracing Agentic Failure from the Flow of Success

  • 类型：arxiv
  • 标识：2607.12747
  • 链接：https://arxiv.org/abs/2607.12747
  • 主分类：agent
  • 形态：method
  • 被引：2
  • 被引来源：Semantic Scholar
  • S2被引：2
  • OpenAlex被引：0
  • 影响力被引：1
  • TLDR：OAT is proposed, which casts this problem as one-class learning with neural controlled differential equations, modeling the dynamical pattern of successful trajectories in latent space, and is shown to be faster than prompting-based baselines and consistently outperforms them in both in-domain and out-of-distribution datasets.
  • OpenAlex ID：W7168405491
  • OpenAlex DOI：10.48550/arxiv.2607.12747
  • DOI：10.48550/arxiv.2607.12747
  • DOI来源：OpenAlex
  • 开放获取：green
  • 开放获取链接：https://doi.org/10.48550/arxiv.2607.12747
  • OpenAlex更新：2026-07-19
  • 待LLM分类：否
  • 标题中文：Tracing Agentic Failure from the Flow of Success
  • TLDR中文：论文提出 OAT，将该问题建模为基于神经受控微分方程的单类学习，在潜空间中刻画成功轨迹的动力学模式；实验表明其比基于 prompt 的基线更快，并在领域内和分布外数据集上均稳定优于基线。
  • 来源文件：
  • /inbox/tom/_candidates/2026-07-16-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]