---
格式版本: 2
标题: "Ignorance or Incompetence? Constructing Knowledge-Gated, Verifiable Tasks for LLM Agents"
原文链接: "https://arxiv.org/abs/2608.30322"
发布日期: "2026-08-31"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Mon, 31 Aug 2026 06:39:27 UTC (439 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-09-02T03:56:40+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-09-02T03:54:32+08:00"
入库时间: "2026-09-01T19:56:40.675Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=Oracle&searchtype=all"
匹配关键词:
  - "AI"
相关厂家:
  - "Oracle"
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 5
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "内容为arXiv论文，主题为LLM Agent知识门控任务构建，与超节点/AI Rack/机柜级AI基础设施完全无关，无相关技术、商业或部署信号。"
AI质检模型: "zj-deepseek-v4-flash"
AI质检时间: "2026-09-02T03:56:56+08:00"
AI主题相关性: 0
AI来源权威性: 5
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 5
AI摘要: "该论文提出一种知识门控的任务构建协议，将任务指令与包含私人惯例、参考表和工具的工件分离，以区分智能体是“不知道”还是“无能力”。在15项校准任务中，某前沿智能体配置提供工件时通过率为68.0%，无工件时为0%；"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-01T23:06:07.773Z"
采集批次: "2026年9月1日23点56分52秒"
采集批次ID: "20260901-235652-761"
去重键: "https://arxiv.org/abs/2608.30322"
---

## Computer Science > Artificial Intelligence

## Title:Ignorance or Incompetence? Constructing Knowledge-Gated, Verifiable Tasks for LLM Agents

Authors:[Hanlin Tian](https://arxiv.org/search/cs?searchtype=author&query=Tian,+H), [Minhao Li](https://arxiv.org/search/cs?searchtype=author&query=Li,+M), [Yu Mi](https://arxiv.org/search/cs?searchtype=author&query=Mi,+Y), [Sihan Zhu](https://arxiv.org/search/cs?searchtype=author&query=Zhu,+S), [Zhao Yang](https://arxiv.org/search/cs?searchtype=author&query=Yang,+Z), [Yuxiang Wang](https://arxiv.org/search/cs?searchtype=author&query=Wang,+Y), [Hongquan Zhu](https://arxiv.org/search/cs?searchtype=author&query=Zhu,+H), [Qiufei Hu](https://arxiv.org/search/cs?searchtype=author&query=Hu,+Q)

[View PDF](https://arxiv.org/pdf/2608.30322) [HTML (experimental)](https://arxiv.org/html/2608.30322v1)

> Abstract:Professional agent tasks often depend on conventions that are absent from public corpora, yet benchmarks rarely control whether an agent has access to those conventions. We introduce a knowledge-gated task-construction protocol that separates a task instruction from a compact artefact containing private conventions, reference tables, and utility operators. Construction-time provenance, byte-identical task instructions across the provided- and withheld-artefact conditions, leak audits, and executable witnesses make dependence on the artefact explicit and testable. Across fifteen calibration tasks, one frontier agent configuration achieves a 68.0% pass rate with the artefact and 0% without it; on one task, a plausible but incorrect artefact also yields 0% across five trials. Deterministic solvers and rule corpora provide exact ground truth for structured tasks, while named criterion-level rubrics support outputs that cannot be checked by a single executable oracle. A configuration-relative calibration screen retains seven tasks satisfying our five-trial empirical knowledge-gating screen. These experiments validate the behavior of the construction protocol; they do not establish that the retained tasks improve post-training. We publicly release part of the task suite and supporting tooling at [this https URL](https://github.com/DatagridsAI/Knowledge-Gated-Task-Construction).

| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| --- | --- |
| Cite as: | [arXiv:2608.30322](https://arxiv.org/abs/2608.30322) \[cs.AI\] |
|  | (or [arXiv:2608.30322v1](https://arxiv.org/abs/2608.30322v1) \[cs.AI\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.30322](https://doi.org/10.48550/arXiv.2608.30322) |

## Submission history

From: Hanlin Tian \[[view email](https://arxiv.org/show-email/4f188651/2608.30322)\]  
**\[v1\]** Mon, 31 Aug 2026 06:39:27 UTC (439 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.30322) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))