最后更新：10/22/2026最后核验：2026-10-15
广告
<p>Deep Review 面向需要在提交或发布前检查复杂文档的研究者、工程师和知识工作者。它把文档分段，分配推理预算，再通过并行智能体检查技术细节、逻辑、数学陈述和相关文献，汇总为供人类判断的反馈。该系统源自 Paper Assistant Tool，已在多个计算机科学会议的作者反馈试点中使用，但不替代正式同行评审。每周包含两次免费审阅，额外审阅需自备 Gemini API Key，费用取决于底层模型和文档用量。提交文档和查看结果需要 Google 账号登录；上传文档会为产品内展示保留最多一年，敏感材料应在上传前评估。</p><ul><li>审阅 PDF 和粘贴文本中的技术错误、薄弱论证与缺失上下文</li><li>检查数学陈述、相关文献和跨章节逻辑一致性</li><li>通过文档分段、并行验证和全局综合生成有依据的审阅反馈</li><li>每周两次免费审阅，额外使用通过自有 Gemini API Key 按量计费</li></ul>
详情
平台
网页版API
功能特性
- Review PDFs and pasted text for technical errors, weak arguments, and missing context
- Check logical consistency, mathematical statements, and related literature across document sections
- Segment documents and combine parallel agent reviews into grounded feedback for human evaluation
- Two free reviews per week, with additional reviews billed through your own Gemini API key
支持语言
en
适合人群
Researchers stress-testing manuscripts before conference submissionEngineers reviewing technical proposals and design-document assumptionsKnowledge workers checking arguments and internal consistency in complex documents
已知限制
- Requires Google Account sign-in to submit documents and view results
- Free access is limited to two reviews per week; extra reviews require a Gemini API key and variable model usage charges
- Uploaded documents are retained for up to one year for display in the product; assess document sensitivity before uploading
- AI feedback supports human review rather than replacing expert judgment or the official conference peer-review process
替代工具
NotebookLM — Source-grounded research and synthesis rather than dedicated technical critique
Claude — General-purpose document analysis and iterative writing in a conversational assistant
Perplexity — Cited web research and source discovery rather than manuscript-level technical review
