arXiv：cs.CL· Dayoon Ko, Jihyuk Kim, Sohyeon Kim, Haeju Park, Dahyun Lee, Gunhee Kim, Moontae Lee, Kyungjae Lee·· 16 小时前
搜索智能体的“幻觉式完成”：约束未验证却给出正确答案高达 48%
When Is Enough Not Enough? Illusory Completion in Search Agents
AI 导读
研究发现，在需要满足多个约束的问题上，LLM 搜索智能体有多达 48% 的正确答案是在仍有约束未验证时就判定任务完成，作者称之为“幻觉式完成”。团队提出 Epistemic Ledger 追踪每轮检索对约束的支撑情况，并测试了从 7B RL 训练模型到前沿 LLM 的 13 个智能体，发现训练和规模只改变验证失败的模式而非消除它。
正文
Abstract: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.
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2602.07549 [cs.AI] |
| (or arXiv:2602.07549v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2602.07549 arXiv-issued DOI via DataCite |
Submission history
From: Dayoon Ko [view email]
[v1]
Sat, 7 Feb 2026 13:50:38 UTC (2,288 KB)
[v2]
Mon, 5 Oct 2026 07:35:12 UTC (1,280 KB)
[v3]
Tue, 13 Oct 2026 22:09:53 UTC (1,220 KB)
来源：arXiv：cs.CL · arxiv.org