← Research

ElicitElicit

Elicit

免费+订阅Free + subscription 学术检索Academic search 结构化抽取Structured extraction 系统性综述Systematic reviews 英文向English-focused partial lifecycle: active 核验Checked 2026-10-22

Elicit 是学术文献的「检索+结构化抽取」工作台：研究问题→论文筛选→抽取表→带引用综述—— 系统性综述流程的 AI 化；每个单元格仍要回原文核。

An academic-literature "retrieve + structured extraction" workbench: research question → paper screening → extraction tables → cited synthesis — a systematic-review pipeline, AI-assisted; every cell still needs checking against the original.

适合与不适合When to use it, when not to

适合：系统性综述/文献调研（英文文献池）、抽取表工作流。不适合：中文文献主打、非学术检索、要全文获取。

Systematic reviews and literature surveys (medical/social-science/ML English literature pools), researchers who want extraction tables rather than abstract-reading. Not for: Chinese literature as the main axis (the academic pool is English-facing), non-academic retrieval, full-text acquisition (it anchors on abstracts and available full-text snippets).

中文生态China availability 国产 / 国内可直接用 / 备案状态 · 快照domestic / direct access / CAC filing · snapshot 2026-10-22

出品属地Origin: 国外出品foreign-made　Elicit（美国）。Elicit (US).
国内使用Direct access: 大陆可直连direct OK　elicit.com 大陆直连实测 200（2026-10-22）。elicit.com measured 200 direct from mainland (2026-10-22).
备案状态CAC filing: 未见于备案名单not in filing list：未见于网信办公示名单（2026-10-22 检索）；该产品未面向大陆公众提供服务，备案不适用。not in the CAC public registry (searched 2026-10-22); the product does not serve the mainland public, so filing does not apply.

备案信息以网信办《生成式人工智能服务已备案信息》公示名单为准（按批次滚动更新，本站只抄编号不解读）；「大陆可直连」是本站从大陆网络直连实测的结果，不是厂商承诺。GitHub 类站点当日实测有波动，口径写在依据里。Filing status is quoted from the CAC public registry (updated batch by batch; this site copies the number and does not interpret it). "Direct access" is this site's own reachability measurement from a mainland network, not a vendor promise. GitHub-hosted sites fluctuated on the measurement day — the note says so.

固定坐标系Fixed coordinate system

8 个维度dimensions, 与同分区其他对象逐项可比each comparable item by item against others in the same section.

模型与开放条件Model & openness: 学术检索系统+LLM 抽取管线：自建论文索引（Semantic Scholar 系学术库为底层之一，官方叙事） +多家 LLM 做抽取/综合（官方叙事，未逐条清点）。
Academic retrieval system + LLM extraction pipeline: a self-built paper index (Semantic Scholar-family academic libraries as one underlying layer, official narrative) + multiple LLMs doing extraction/synthesis (official narrative, not inventoried item by item).
运行位置Where it runs: 网页（elicit.com）为主入口；文献检索与抽取生成都在厂商云端完成，未见官方声明的本地形态。
Web (elicit.com) as the main entry; literature retrieval and extraction generation both happen on the vendor's cloud; no officially stated local form seen.
本地文件Local files: 项目/文献库存账户侧云端；导出（CSV/综述文档）归本地。未见官方声明的本机数据面。
Projects/literature libraries stored on the account-side cloud; exports (CSV/review documents) count as local. No officially stated local data surface seen.
关机后的任务Tasks after shutdown: 检索即时；系统性综述类任务（多轮筛选/抽取表）是任务线；无「关机续跑」概念。
Retrieval is instant; systematic-review style tasks (multi-round screening/extraction tables) are a task line; no "keep-running-after-shutdown" concept.
工具与扩展Tools & extensions: 文献检索+结构化抽取是主轴：按研究问题检索论文、筛选（纳入/排除标准面）、抽取表（样本/方法/结论列）、综述综合（带引用）；「找相似论文」线（官方产品面）。
Literature retrieval + structured extraction is the main axis: searching papers by research question, screening (inclusion/exclusion criteria surface), extraction tables (sample/method/conclusion columns), review synthesis (with citations); a "find similar papers" line (official product surface).
上下文与记忆Context & memory: 以项目（project）组织文献工作；跨会话记忆未核验到官方声明。
Literature work organized by project; cross-session memory not verified to an official statement.
权限与限制Permissions & limits: 账号+订阅体系（D9：价格不采）；免费档检索/抽取额度限制（未逐条核验）；训练用途开关未逐条核验。
Account + subscription system (D9: pricing not collected); free-tier retrieval/extraction quota limits not verified item by item; the training-use switch not verified item by item.
适合什么任务Best for: 适合：系统性综述/文献调研（医学/社科/ML 等英文文献池）、要「抽取表」而非「读摘要」的研究者。不适合：中文文献主打（学术库英文向）、非学术检索、要全文获取（它锚定摘要与可得全文片段）。
Fits: systematic reviews/literature surveys (medicine/social sciences/ML and other English literature pools), researchers who want "extraction tables" rather than "reading abstracts". Does not fit: Chinese-literature-first (the academic libraries are English-leaning), non-academic retrieval, wanting full-text access (it anchors abstracts and available full-text snippets).

头号误解Top misconception

「AI 帮你读文献」≠「AI 结论可信」——抽取表的每个单元格要回原文核，尤其数值类"AI reads papers for you" is not "AI conclusions are trustworthy" — every cell of the extraction table needs checking against the original, especially numbers
检索覆盖取决于底层学术库——不在库里的论文检索不到，系统性综述要多库交叉Retrieval coverage depends on the underlying academic database — papers outside it cannot be found; systematic reviews need multi-database cross-checks
elicit.com 对 curl 返回 CF 403——本站核验方法受限如实标注，不是站点问题elicit.com returns a Cloudflare 403 to curl — our verification method is limited and noted honestly; it is not a site problem

未知项清单Known unknowns

底层学术库清单与覆盖（官方叙事 Semantic Scholar 系，未逐条核验）
当前使用的 LLM 清单（未逐条清点）
ToS 细节（CF 403 未能直抓）
训练用途开关（未逐条核验）

The underlying academic-library list and coverage (official narrative says Semantic Scholar-family, not verified item by item)
The LLM list currently in use (not inventoried item by item)
ToS details (CF 403 blocked direct fetch)
The training-use switch (not verified item by item)

证据来源Evidence sources

判断可回到以下一手源复核。本站核验日Every judgement can be rechecked against the primary sources below. Verified on 2026-10-22，内容更新日content updated on 2026-10-22。

类型Type	名称Name	链接Link
official	Elicit · 官网（产品与能力叙事）Elicit · official site (product and capability narrative)	https://elicit.com
official	Elicit · Terms（官方条款，直抓 CF 403，2026-10-22）Elicit · Terms (official terms; direct fetch hit CF 403, 2026-10-22)	https://elicit.com/terms
changelog	Elicit · Blog（官方更新线，changelog）Elicit · Blog (official update line, changelog)	https://elicit.com/blog

实测记录Measured runs

本档未跑检索质量评测；结论只覆盖产品能力与来源披露。

实测建议（三条，都指向官方可读页没给的东西）：

抽取表数值类字段的准确率（对照原文抽样）
检索召回 vs 手工 PubMed 检索的差集
免费档额度边界（检索/抽取次数）

This site has not completed measured runs. The coding-oriented test protocol is at protocols/coding/_protocol.md (the coding task suite does not apply to creation products).

Suggested measurements (three, all pointing at things the official readable pages do not give):

Accuracy of numeric fields in extraction tables (sampling against the originals)
The set difference between retrieval recall and a manual PubMed search
The free-tier quota boundary (retrieval/extraction counts)