能動性克制：智能體是否知道何時該停止而非行動？

摘要

LLM 代理被期望在多轮交互中行动，利用搜索、浏览界面和终端工具来完成用户目标。然而，并非每个目标在可用环境中都能被充分指定或实现。在这种情况下，可靠的代理应认识到进一步交互不太可能有所帮助，从而放弃额外的工具调用。我们定义了“智能体弃权”（Agentic Abstention）问题，即决定代理在不确定性下何时应停止行动的问题。与标准的 LLM 弃权（通常作为单轮的“回答或弃权”决策进行评估）不同，智能体弃权是一个序贯决策问题：代理在每一轮中可以选择回答、弃权或收集更多信息，而弃权的必要性可能只有在与环境交互后才变得清晰。我们在网络购物、终端环境和问答三类场景中研究了这一问题，评估了 13 个 LLM 代理系统和 2 个代理框架，涉及超过 28,000 项任务。我们的结果表明，主要挑战不仅在于代理能否弃权，更在于它们何时弃权。有些代理在应该弃权时从未弃权，而另一些则是在多次不必要的交互之后才弃权。这一差距在那些指令看似可行但环境后来揭示其不可行（例如，没有有效结果匹配指令）的任务中尤为显著。我们进一步发现，模型规模、推理能力和代理框架对弃权行为有不同影响，有时更大的模型或能力更强的模型在及时弃权方面表现更差。最后，我们引入了 CONVOLVE，一种通过将完整的交互轨迹提炼为可复用的停止规则来改进智能体弃权的上下文工程方法。在 WebShop 上，CONVOLVE 在不更新模型参数的情况下大幅提升了及时弃权能力，将 Llama-3.3-70B 的及时召回率从 26.7% 提高到 57.4%。我们的数据集和代码可在 https://lhannnn.github.io/agentic-abstention 获取。

English

LLM agents are expected to act over multiple turns, using search, browsing interfaces, and terminal tools to complete user goals. Yet not every goal is well specified or achievable in the available environment. In such cases, a reliable agent should recognize that further interaction is unlikely to help and abstain from additional tool calls. We define Agentic Abstention, the problem of deciding when an agent should stop acting under uncertainty. Unlike standard LLM abstention, which is usually evaluated as a single-turn answer-or-abstain decision, agentic abstention is a sequential decision problem: an agent can answer, abstain, or gather more information at each turn, and the need to abstain may only become clear after interacting with the environment. We study this problem across web shopping, terminal environments, and question answering, evaluating 13 LLM-as-agent systems and 2 agent scaffolds on more than 28,000 tasks. Our results show that the main challenge is not only whether agents can abstain, but also when they abstain. Some agents never abstain when they should, while others do so only after many unnecessary interactions. This gap is especially large on tasks where the instruction appears feasible until the environment reveals otherwise (e.g., no valid result matches the instruction). We further find that model scale, reasoning, and agent scaffolding affect abstention in different ways, where larger or more capable models sometimes perform worse at timely abstention. Finally, we introduce CONVOLVE, a context engineering method for improving agentic abstention that distills full interaction trajectories into reusable stopping rules. On WebShop, CONVOLVE substantially improves timely abstention without updating model parameters, raising Llama-3.3-70B's timely recall rate from 26.7 to 57.4. Our dataset and code are available at https://lhannnn.github.io/agentic-abstention