arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.07549cs.AIcs.CL
收藏

何时足够并不足够?搜索代理中的错觉完成

When Is Enough Not Enough? Illusory Completion in Search Agents

Seoul National University(首尔国立大学);LG AI Research(LG人工智能研究);University of Illinois Chicago(伊利诺伊大学芝加哥分校)

Dayoon Ko, Jihyuk Kim, Sohyeon Kim, Haeju Park, Dahyun Lee, Gunhee Kim, Moontae Lee, Kyungjae Lee

AI总结:

研究发现搜索代理在多约束问题中常出现错觉完成,通过引入认知账本框架,显式跟踪约束状态可有效减少未经验证答案并提升准确性。

摘要:

最近的搜索代理利用多轮推理和搜索工具,在多跳和长周期基准上实现了强劲的性能。然而,仍不清楚这些代理是否能够可靠地在这些问题中跟踪、验证和维护多个条件,以确保所有要求都被满足。我们研究了这种能力在多约束问题中的表现,其中有效答案必须同时满足多个约束条件。我们发现错觉完成经常发生,即代理认为任务已完成,尽管存在未解决或违反的约束,导致答案未经验证。为了诊断这种行为,我们引入了认知账本,一种评估框架,它在整个多轮推理过程中跟踪每个约束的证据支持和代理的信念。我们的分析揭示了四个反复出现的失败模式:空声明、被忽视的反驳、停滞和过早退出。受这些发现的启发,我们研究了在执行过程中显式跟踪约束状态是否能够缓解这些失败,通过LiveLedger,一种推理时的跟踪器。这种简单的干预措施一致地提高了性能,显著减少了未经验证的答案(减少多达26.5%),并在多约束问题上提高了整体准确性(提高多达11.6%)。

英文摘要:

In agentic search, an LLM agent searches the web, reads the pages it finds, and decides what to look for next before returning an answer. But can we trust an answer simply because the agent returns it? Often not, and even a correct answer can be a lucky guess: on questions with several constraints, we find that agents conclude the task is complete while a constraint remains unverified in up to 48% of their correct answers. We call this illusory completion. To see how it arises, we introduce the Epistemic Ledger, which tracks at every turn what the retrieved pages establish about each constraint and what the agent claims. Across 13 agents, from 7B RL-trained models to frontier LLMs, training and scale raise accuracy but change the pattern of verification failures rather than eliminating them: constraints may be left unchecked, assumed without support, or retained despite refuting evidence. To measure what agents lose without tracking their constraints, we show them each constraint's state, approximated by LiveLedger, a lightweight 4B tracker. Agents then answer 4.4-16.1 points more questions correctly, suggesting that on their own, they may not track what they have verified and what remains.

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。

↑