Topics & Evidence选题与证据

ISEF, Read from the Record

从记录出发的 ISEF 选题

330 scoped topics across 22 tracks, and an honest reading of who actually wins.

22 个赛道、330 则候选课题，以及一份关于「谁真的拿奖」的诚实解读。

Two things live here. The first is a set of candidate topics, each built by the same move: take a recent winning project and ask the question it never asked about its own validity. The second is the record those topics are checked against — every ISEF finalist from 2022 to 2026, 77 countries, 6,755 projects, read with grand awards and special awards counted separately.

这里放着两样东西。一是候选课题，每一条都出自同一个动作：拿一个近届获奖作品， 问它没问的那个关于自身有效性的问题。二是这些课题所对照的记录—— 2022 至 2026 年全部 ISEF 入围项目，77 个国家与地区、6,755 个项目， 并且把大奖与专项奖分开计算。

That separation matters more than it sounds. Combine the two kinds of award and American finalists lead at 40.3% against mainland China's 38.0%. Count grand awards alone — the ones ISEF's own subject panels hand out with everyone competing on the same floor — and mainland China leads at 35.8% against 31.2%. The difference is that 504 of the 931 American special-award entries are US university scholarships or government-agency prizes an international student cannot claim.

这个区分比听上去重要。把两类奖合并，美国选手 40.3% 领先中国大陆的 38.0%； 只算大奖——由 ISEF 学科评委团评出、全体入围者同台竞争的那一类—— 中国大陆 35.8% 反超美国 31.2%。差别在于美国 931 条专项奖里有 504 条是 美国大学奖学金或政府机构奖，国际学生拿不到。

Where to begin从何处开始

Seven ways in七种进入方式

The Data数据

What the record says从数据里读 ISEF

Five years of finalists across the world, China and the United States — with a switch that separates grand awards from special awards, and confidence intervals on every small sample.全球、中国、美国三个视角的五届入围记录——可切换大奖与专项奖口径，小样本一律给置信区间。

Open the analysis →查看数据分析 →

The Tracks赛道

22 categories, 156 subcategories22 个类别，156 个子类别

The official definition of every category and subcategory, what each track’s five-year record looks like, and the projects that actually took first place in it.每个类别与子类别的官方定义、该赛道五届的完整记录，以及真正在其中拿下一等奖的作品。

Open the tracks →查看赛道 →

The Map地图

17,564 projects, arranged by subject17,564 个项目，按内容排布

Thirteen seasons of public abstracts laid out by what they are about rather than which category judged them, with 10,632 measured content links you can follow project by project.十三届公开摘要，按研究内容而不是评审类别排布，10,632 条算出来的内容关系可以逐个项目跟下去。

Open the universe →打开项目宇宙 →

China 20262026 中国

Who got through, and from where谁晋级了，从哪个赛区

All 51 Greater China projects at the 2026 finals, grouped by the affiliated fair each came through — because the quota belongs to the fair, not to the country.2026 年决赛上全部 51 个大中华区项目，按各自晋级的附属赛区分组——名额属于赛区，不属于国家。

See the fairs →查看各赛区 →

Candidate Topics候选课题

330 scoped topics330 则候选课题

Fifteen per track across all 22 ISEF categories — each with acceptance criteria, a novelty boundary, an equipment-and-time gate, and the part you still have to supply yourself.22 个类别各 15 则——每则都有验收标准、新颖性边界、器材与时间门槛，以及「你自己的那一点」。

See the topics →查看候选课题 →

Tooling工具

Skill assistants技能助手

Twelve installable assistants covering topic search, judge simulation, research plans, abstracts, posters, interview drill and compliance forms.十二个可安装的助手，覆盖选题检索、评委模拟、研究计划、摘要、海报、答辩训练与合规表格。

Browse the skills →浏览技能助手 →

Method方法

How this was built这些是怎么做出来的

The corpus, the scoring, the boundaries — including what the data cannot tell you and where a label is a judgement call rather than a fact.语料、评分、边界——包括数据不能回答什么，以及哪些标签是判断而非事实。

Read the method →阅读方法说明 →

Three things the data settles数据能定的三件事

Before you pick a topic在选题之前

Headcount is quota, not talent. A state or a country sends roughly the same number of finalists every year because allocation is written into each affiliated fair's agreement, not earned afresh each season. Florida sent 87, 84, 87, 92, 86 across five years. Taiwan sent exactly twelve, five years running. Reading "that place had a lot of finalists" as "that place is strong" gets the causality backwards.

人数是配额，不是实力。一个州、一个国家每年送出的人数几乎固定，因为名额写在 每个附属赛区的协议里，不是每季重新挣来的。佛罗里达五年是 87、84、87、92、86； 台湾连续五年整整 12 项。把「那边入围的人多」读成「那边学生强」，因果是反的。

Getting in and winning are different problems. Florida sends the most finalists in the country and wins below the national rate; Arizona tops the combined award rate at 66.3% and falls to 27.6% on grand awards alone. Which school you enter through changes your odds of getting in; it does not change what the judges ask you.

进决赛和拿奖是两个问题。佛罗里达入围人数全美第一，获奖率低于全国平均； 亚利桑那合并口径 66.3% 全美第一，只算大奖掉到 27.6%。从哪个赛区进去， 影响的是进不进得去；影响不了评委问你什么。

Some tracks are nearly empty. Chinese finalists crowd into engineering, physics, robotics and mathematics, and are close to absent from computational biology and biochemistry — 0.8% and 0.4% against the American 6.4% and 3.2%. Crowding is not a reason to switch fields, but it is worth knowing which room you are walking into.

有些赛道几乎空着。中国选手扎堆在工程、物理、机器人与数学， 在计算生物与生物化学上接近缺席——0.8% 与 0.4%，而美国是 6.4% 与 3.2%。 拥挤不是换方向的理由，但值得知道自己走进的是哪一间屋子。

Open the full analysis →查看完整数据分析 →