Yau Awards Archive 2020 — 2025

CHAPTER EIGHT第八章

丘成桐中学科学奖面试答辩准备 Preparing for the Oral Defense

08 · From the Whitepaper, v22.0 · May 2026 摘自白皮书 v22.0 · 2026 年 5 月


论文通过初筛、函评、桌评之后,丘奖剩下的两轮评审都是当面进行的:分赛区决赛(往届称“半决赛”)和全球总决赛。两轮都是面试答辩,每队 30 分钟,一半时间讲,一半时间答。组委会在奖项设置中写明,每年实际授予的奖项数量“以本年度参赛作品质量和学生答辩表现的综合评判协商确定”。也就是说,到了这两轮,论文写得好只是入场券,最终名次要看你能不能在评委面前讲清楚、答得住。

本章先把官方公布过的答辩规则逐条整理出来(第 1 节),再讨论两轮答辩的差异、演示文稿的制作、讲述与时间控制、评委问答的准备方法,最后给出赛前训练时间表和自查清单。凡是官方规则,我们都注明出处;凡是经验建议,我们都明确标为建议。两者请区分对待:规则是必须遵守的,建议可以按自己项目的情况取舍。

答辩赛制:官方规则梳理

本节依据以下官方来源整理:丘奖官网“常见问题解答”(Q12、Q20,核实于 2026-08-31)1、2026 年参赛规则与参赛日程2、2023–2025 年中国内地赛区半决赛入围名单公告3、2020 与 2021 年《总决赛学生手册》4,以及 2024 年总决赛参赛通知5。历年细节会调整,一切以当季组委会和各赛区的正式通知为准。

两轮答辩一览

丘奖两轮面试答辩对比(中国内地赛区)
项目 分赛区决赛(半决赛) 全球总决赛
2026 年时间 10 月 1 日至 11 月 2 日分赛区评审期内,由各赛区另行通知。往届实际举办日期:2023 年 10 月 28/29 日,2024 年 11 月 2/3/4 日,2025 年 11 月 8/9 日 12 月 5 日至 6 日,清华大学;12 月 6 日另有科学论坛
时长结构 每队 30 分钟:15 分钟 PPT 演示 + 15 分钟评委问答 每队不超过 30 分钟:指导老师介绍 1–2 分钟 + 学生 PPT 演示约 13–14 分钟 + 评委问答 15 分钟
语言 官方 FAQ:总决赛之前的阶段,中英文没有强制要求(具体以赛区通知为准) 全英文:最终版论文、答辩 PPT、现场答辩均须使用英文
参加人员 学生参加,指导老师不参加 学生和指导老师均须参加;指导老师介绍完毕后离开答辩室
评审方 分赛区评审委员会 国际评审委员会
入围到答辩的间隔 往届半决赛名单公布后约 9–11 天即答辩(2023:10/18 公布,10/28 答辩;2024:10/22 公布,11/2 答辩;2025:10/30 公布,11/8 答辩) 往届总决赛入围名单公示后约 4 周(2025:11/11 公示,12/6 总决赛)

表 1 中最值得注意的是最后一行。半决赛名单公布到答辩往往不到两周,如果等名单出来才开始做 PPT、练讲稿,时间非常紧张。第 6 节会给出一个从提交论文就开始准备的时间表。

总决赛答辩室规则

以下规则综合自 2020、2021 年《总决赛学生手册》和 2024 年总决赛通知,三份文件在核心规则上一致:

  1. 计时从进门开始。参赛队员一进入答辩室,计时即正式开始,整场不超过 30 分钟。拷贝文件、打开 PPT、调试设备的时间都算在内。

  2. 讲述可以一人主讲,也可以多人分工;回答问题时各队员可以自由作答。手册建议每位学生都有所发言。

  3. 答辩内容的官方要求:需要对问题提出的背景、关键背景材料、参赛队伍的所有原创内容以及队员间的分工进行阐述。这一条直接对应评审标准中的“口语表达的学术规范性”和“团队合作性”,请把它当成 PPT 必须覆盖的内容清单。

  4. 指导老师环节(2022 年起设立):指导老师进入答辩室,用 1–2 分钟简要介绍指导情况并确认学生身份,然后离开到候场室等候。2024 年通知列明的介绍要点包括:学生是否付费接受指导;是否还有其他指导老师;是否有培训机构参与;学生的参与程度与主要贡献。中学指导老师还须确认参赛学生是本校学生。

  5. 其他人员:家长、领队等其他人员不得进入答辩室(2024 年通知规定,指导老师因特殊情况无法到场时,可由一名领队或家长代为出席,须事先向组委会说明)。

  6. 候场与签到:须在各自答辩时间前 30 分钟到达候场室,未按时到达视为放弃比赛;携带身份证或护照(港澳学生携带通行证)及学生证;候场区手机静音,可以小声练习,不得影响他人。

  7. 设备:2024 年通知称,答辩室提供 Windows 10 电脑、翻页笔等设备,PPT 应为 16:9 比例;不建议使用个人电脑,如使用个人电脑或现场更新 PPT,调试时间计入答辩时间。

  8. 材料提交:公示期结束后,入围团队须在规定截止时间前(2024 年为 11 月 30 日 24:00)通过报名系统提交英文最终版论文和答辩 PPT,逾期不收。如果最终版论文与公示版相比有增补或修改,须把修改部分详细列出,单独上传到“其他材料”栏。

  9. 科学金奖面试:总决赛次日上午另有科学金奖答辩,由各学科金奖团队参加。往届名单在总决赛当晚通知,请保持手机畅通。

提醒:2021 年总决赛为线上线下结合,2022 年总决赛首次全部线上进行。
如果当季采用线上答辩,摄像头、网络、共享屏幕、监考要求等细节
请以当季通知为准,并提前完整测试一遍。

评审标准中与答辩直接相关的条目

在第 1 章各学科评审标准中,有两条是专门针对答辩的。第一条是口语表达的学术规范性,六个学科的表述相同:答辩“需要说清楚问题的背景,关键背景材料,以及参赛队伍的所有原创内容”。第二条是团队合作性,各学科的要求有松有紧:

  • 化学、生物、计算机:如是团队参赛,每位队员都必须口头报告部分内容,并回答评委所提问题。计算机奖还特别写明“避免搭车参赛”。

  • 数学、物理、经济金融建模:每个参赛学生都应该“有组织地代表其所在的队伍发言”。

此外,化学奖评审标准要求研究成果必须由参赛选手在指导教师指导下完成,物理、化学、生物、计算机、经济金融建模五个学科都写明评委“特别关注参赛团队的学术道德和诚信”。答辩是评委核实这一点最直接的机会。2024 年总决赛的官方总结列出的考核维度为:创新性、原创性、学术规范性、团队协作能力、发展潜力、学术道德诚信。

两轮答辩的差异与备战重点

两轮答辩形式相近,但侧重点不同,准备时要分开考虑。

分赛区决赛:决定能否进入总决赛。这一轮的评委来自分赛区评审委员会,面对的是本赛区所有入围团队。评委通常已经读过论文,问答会集中在论文里说得不清楚的地方:方法是否站得住、数据从哪里来、哪些工作是学生自己做的。语言方面,官方 FAQ 写明总决赛之前中英文没有强制要求,但具体用什么语言答辩、PPT 用什么语言,要看本赛区的通知。如果论文是用英文写的,建议提前问清楚,并同时准备中英两套讲稿。

全球总决赛:决定奖项等级。总决赛由国际评审委员会主持,评委中有不少是海外学者(参见第 4 章评委背景)。与分赛区决赛相比有三点变化:第一,全程英文,最终论文和 PPT 都要换成英文版;第二,指导老师要到场介绍指导情况,这段介绍会被评委拿来和论文、答辩内容互相对照;第三,入围团队的水平更接近,评委会追问得更深,数学奖评委要求现场推导关键步骤是常见情形。

导师介绍、致谢页、指导老师信息表要一致。这是很多团队忽略的一点。指导老师在答辩室口头介绍的指导情况(是否有偿、有无其他导师、有无机构参与、学生贡献),与论文致谢页的说明、报名时提交的《指导老师信息表》以及学术诚信声明,说的是同一件事。赛前请学生和指导老师把这几份材料一起过一遍,确保说法一致。说法不一致本身就会引起评委对学术诚信的疑问。

演示文稿(PPT)制作

时间预算与页数

半决赛有 15 分钟,总决赛扣除导师介绍后约 13–14 分钟。按照一般学术报告每页约 1 分钟的节奏(经验值),正文控制在 12–15 页比较合适。页数太多会导致语速过快,或者讲到后面被迫跳页,而最后几页恰好是结果和结论,是最不应该被压缩的部分。

推荐结构

表 2 是一个按 13 分钟设计的结构建议,适用于大多数学科。其中“原创贡献”和“团队分工”两页对应上文的官方答辩要求,建议不要省略。

答辩 PPT 推荐结构(按约 13 分钟设计,仅供参考)
部分 内容要点 建议时长
标题页 题目、队员、学校;用一句话说出研究问题 0.5 分钟
研究背景 这个问题为什么值得研究;已有研究做到哪一步,缺什么。只讲和你的工作直接相关的背景 1.5 分钟
原创贡献 单独一页,明确写出“本研究的原创工作是……”,把背景材料与自己的工作分开 1 分钟
研究方法 模型、推导、实验设计或算法;说明为什么选这个方法而不是其他方法 3 分钟
结果 用具体数字、带误差的图表、与对照组或基线的比较呈现 3.5 分钟
讨论与局限 结果说明了什么;在哪些条件下不成立;还有什么没做 1.5 分钟
结论与展望 重申贡献;下一步可以做什么 1 分钟
团队分工与致谢 每位队员负责的部分;指导老师的作用;AI 工具使用情况(如有,参见第 7 章) 1 分钟
备用页 放在结尾之后,不在正式讲述中出现,用于回答预计会被问到的问题 不计入

制作细节

  1. 比例与兼容性。按 16:9 制作。答辩室提供的是 Windows 电脑,用 Keynote 或 Mac 版软件制作的文件,在 Windows 上可能出现字体替换、公式错位、动画失效。定稿后请在一台 Windows 电脑上完整播放一遍,并同时准备一份 PDF 版本作为备份。

  2. 每页只讲一件事。页面标题直接写结论,例如写“Method B reduces error by 38%”,而不是只写“Results”。评委扫一眼标题就知道这页要说明什么。

  3. 图比文字重要。坐标轴要有标签和单位,字号在投影上看得清;实验数据标出样本量和误差;与基线比较时,把自己的方法用颜色突出。

  4. 加页码,给公式和图编号。评委提问时常说“回到第 7 页”“你第二个公式”,有页码和编号能节省宝贵的问答时间。

  5. 少用动画。逐条飞入的动画会拖慢节奏,也容易在换电脑后出错。

  6. 备用页要准备充分。把完整推导、额外的实验、消融实验、数据来源说明、参数设置表放在备用页里。评委问到时直接翻出来,比口头描述有说服力得多。

  7. 与论文保持一致。PPT 中出现的数字、图表、术语要和最终提交的论文一致。如果论文公示后又补做了实验,要按规定在“其他材料”中说明修改内容,并在答辩中主动说明哪些是新增结果。

讲述与时间控制

团队如何分工讲述

多人团队建议按研究内容分工,而不是平均切分页数。一个常见的做法是:谁主导了哪部分工作,就由谁讲哪部分,问答时也由他主答该部分的问题。例如三人团队可以一人讲背景和问题提出,一人讲方法,一人讲结果和讨论。这样做有两个好处:一是讲述者对内容最熟,二是评委能直接看到每个人的贡献,这正好对应评审标准中的团队合作性。

交接处要事先设计好过渡句,比如“Now my teammate will present the experimental results.”,避免出现冷场或抢话。

控制时间

  • 按比规定时间少 1 分钟来练。半决赛练到 14 分钟以内,总决赛练到 12–13 分钟以内。进门、问好、打开文件都在计时之内,现场讲述通常也比练习时慢。

  • 设置时间检查点。例如讲到方法结束时应不超过 6 分钟。练习时记录每一部分的实际用时,超时的部分删内容,不要靠加快语速。

  • 准备“压缩版”。如果现场被提前打断或时间不够,知道哪几页可以一句话带过、哪几页必须讲完。

不要背稿

逐字背诵的讲稿,一旦被评委中途打断或忘记一句,很容易卡住,听起来也像在念稿。建议每页只记住两三个要点和一个关键句,其余用自己的话讲。一个检验方法是:关掉 PPT,能否用 3 分钟口头讲清楚整个研究的问题、方法、结果和结论。能做到,说明你真正掌握了内容。

对总决赛来说,英文表达是额外的难点。一个实际的建议是:把自己研究中反复出现的专业术语(方法名称、指标、关键概念)整理成中英对照表,每个词都大声读到流利为止。在自己的核心术语上结巴,评委会怀疑你对工作的熟悉程度。

评委问答

问答环节通常决定了评委对团队的最终判断。评委已经读过论文,提问往往不是为了考定义,而是为了确认三件事:这项工作是不是你做的,你是否理解自己用的方法,你的结论是否站得住。

常见问题类型

以下分类综合自评审标准和往届参赛者的公开经验,按类型整理,每类附中英文例题。请针对自己的项目,把每一类至少写出三个具体问题并练习回答。

评委问题类型与例题
类型 例题
研究动机 为什么选这个问题?它解决了什么实际或理论问题?
Why is this problem worth studying?
方法选择 为什么用这个方法,而不是更简单或更常用的方法?
Why did you choose this approach over X?
严谨性 怎么排除偶然性、过拟合或混杂因素?对照组 / 基线是什么?
How do you know this result is not due to chance?
机制解释 这个结果为什么会出现?请解释原因,而不只是描述现象。
Can you explain why it works?
原创与分工 哪部分是你们自己做的?每个人具体负责什么?导师帮了什么?
Which part of this work is your own contribution?
局限与延伸 这个方法在什么情况下不适用?如果再给你半年,你会做什么?
What are the limitations? What would you do next?
文献核查 你引用的第 X 篇文献,核心结论是什么?
Can you summarize reference [X] in one sentence?
AI 使用 研究中用了哪些 AI 工具?用在哪个环节?不用 AI 能否完成?(详见第 7 章)

各学科的追问侧重

结合第 4 章对各学科评委背景的分析,各学科的追问通常有以下侧重(经验总结,不代表每场答辩都会出现):

  • 数学:要求在白板上重现关键引理或定理的证明;追问某个假设条件为什么必要,去掉会怎样。关键推导必须能脱稿写出。

  • 物理:量纲检查、边界条件、近似的适用范围;实验项目追问误差的主要来源和测量方法。带实物或实验装置照片、视频会有帮助。

  • 化学:产物如何表征确认(NMR、XRD、质谱等);为什么选这条合成路线;计算化学项目追问泛函和基组的选择理由。

  • 生物:样本和数据来源、伦理审批与知情同意;对照组设计、重复次数、批次效应;实验是在哪个实验室由谁完成的。

  • 计算机:数据集如何划分训练集、验证集、测试集;基线比较是否公平;模型结构和超参数的具体取值及理由;提升幅度是否显著。

  • 经济金融建模:因果识别策略(DID、IV、RDD 等)和关键假设是否成立;有没有做稳健性检验和安慰剂检验;相关性是否被当成了因果。

回答的基本原则

  1. 先给结论,再给依据。第一句直接回答问题,然后用一两句说明理由或证据。避免从背景讲起、绕一大圈才回到问题。

  2. 用具体数字和细节。“效果提升了”不如“在测试集上准确率从 0.71 提高到 0.79”。具体的回答本身就说明你熟悉自己的工作。

  3. 不会就说不会,但给出思路。遇到答不上的问题,不要编造,也不要只说“I don’t know”。可以说明目前没有做这部分,并提出如何验证,例如 “We haven’t tested that yet, but we would check it by …”。评委更看重诚实和思考方式。

  4. 没听清就先确认。英文答辩中听不清或不确定题意很常见,可以先复述问题:“Just to make sure I understand, are you asking about …?” 这比答非所问好得多。

  5. 不把责任推给导师。“这是老师让我们这样做的”会让评委认为学生不理解自己的研究设计。即使某个选择来自导师建议,也要能说清楚理由。

  6. 承认局限,不过度声称。评委指出的问题如果确实存在,坦率承认并说明可能的改进方向,比辩解更能赢得信任。结论只说到证据能支持的程度。

  7. 团队内事先约定谁答。按分工确定每类问题的主答人,其他队员可以补充,但不要抢答或互相矛盾。评委点名提问时,被点名的人必须自己回答。

常用英文表达

开场:   Good morning. We are ... from ... Our project is about ...
过渡:   Next, my teammate ... will explain our method.
确认题意:Could you please repeat the question?
          Just to make sure I understand, are you asking about ...?
争取思考:That's a good question. Let me think for a moment.
承认局限:You're right, this is a limitation of our current study.
          We plan to address it by ...
未做的工作:We haven't tested that yet, but we would check it by ...
引用备用页:We have a backup slide on this. May I show it?
结束:   Thank you. We are happy to take your questions.

赛前训练计划

时间表

由于半决赛名单公布到答辩往往只有 9–11 天,建议在 9 月 15 日提交论文后就开始准备答辩,而不是等入围通知。表 4 以 2026 年赛程为例(具体日期以官方通知为准)。

答辩准备时间表建议(以 2026 年赛程为例)
时间段 准备内容
9 月 16 日 – 9 月 30 日(材料审核期) 完成第一版 PPT 和讲稿;整理问题清单;数学、物理项目练习关键推导的板书
10 月上中旬(等待入围通知) 至少进行 2 次完整的计时模拟答辩;请不熟悉课题的老师或同学扮演评委;根据问答暴露的问题补备用页
半决赛名单公布后(约 10 天) 确认本赛区的答辩时间、地点、语言和设备要求;再进行 2–3 次模拟答辩;在答辩所用的电脑环境上测试文件
半决赛结束后 记录评委提出的所有问题,逐条反思哪些没答好;开始把论文和 PPT 改为英文
总决赛公示期(11 月上中旬) 完成英文论文和英文 PPT;进行英文模拟答辩;与指导老师核对导师介绍内容与致谢页、信息表是否一致
提交截止前(往届为 11 月底) 按时提交英文终稿和 PPT;如论文较公示版有改动,列出修改清单上传到“其他材料”
总决赛前一周 每天至少完整讲一遍;准备证件、U 盘(PPT 与 PDF 各一份);确认交通和到场时间

如何做模拟答辩

  1. 严格按正式流程。从敲门进场开始计时,15 分钟讲、15 分钟问,中途不暂停。总决赛模拟时,请指导老师也按正式要求做 1–2 分钟介绍。

  2. 评委最好不是指导老师本人。指导老师太熟悉课题,很难问出外行评委的问题。可以请相关学科的其他老师、大学生或研究生来提问,至少包括一位非本学科背景的人。

  3. 录像回看。重点看三点:有没有超时;有没有对着屏幕念 PPT;回答问题时第一句话是否直接回应了问题。

  4. 建立问题库。把每次模拟中被问到的问题和自己的回答要点记下来,答得不好的问题反复练习。半决赛的真实问题尤其要保留,总决赛常会被问到同一类问题,而且追问得更深。

答辩自查清单

答辩前请逐条核对,任何一项做不到都应抓紧补救:

  1. 我们已确认本轮答辩的时间、地点、语言、设备和提交要求,以当季正式通知为准。

  2. 我们的 PPT 为 16:9 比例,已在 Windows 电脑上完整播放测试,并准备了 PDF 备份。

  3. PPT 中有单独的页面说明研究背景与原创内容的区别,以及每位队员的分工。

  4. 每位队员都有讲述部分,并且都能独立回答自己负责部分的问题。

  5. 完整讲述时间控制在规定时间内,并留有至少 1 分钟余量。

  6. 我们能脱离 PPT,口头讲清研究问题、方法、结果和结论。

  7. PPT 和讲述中的所有数字、图表与最终提交的论文一致。

  8. 我们能用一句话说出所引用的每一篇主要文献的核心结论。

  9. (数学、物理等)我们能在白板上独立写出关键推导,不需要看资料。

  10. 我们准备了覆盖常见问题的备用页,并知道每张备用页在第几页。

  11. 如果使用过 AI 工具,我们能说清楚用了什么、用在哪个环节、哪些工作是自己完成的,说法与致谢页一致。

  12. 指导老师的介绍内容与论文致谢页、指导老师信息表、学术诚信声明一致。

  13. (总决赛)英文论文、英文 PPT 已在截止时间前提交;论文较公示版的改动已列出并上传。

  14. 证件、学生证、U 盘已准备好,并计划在答辩前至少 30 分钟到达候场室。

结语。答辩的 30 分钟,是评委唯一一次能当面了解你和你的研究的机会。评委并不期待高中生的报告像学术会议一样完美,他们想确认的是:这项研究确实是你做的,你真正理解它,并且能诚实地说出它的价值和局限。做到这三点,比漂亮的 PPT 和流利的英文更重要。


  1. https://www.yau-awards.com/page-assessment.html↩︎

  2. https://www.yau-awards.com/page-rule.html;https://www.yau-awards.com/show-86-57.html,发布于 2026-04-16。↩︎

  3. https://www.yau-awards.com/show-86-34.html、https://www.yau-awards.com/show-86-42.html、https://www.yau-awards.com/show-86-50.html。↩︎

  4. 官网“文件下载”页:https://www.yau-awards.com/lists-download.html。↩︎

  5. 2024 年总决赛通知的英文整理见 https://en.embarkchina.org/blog/1784.html;官方总结见 https://www.yau-awards.com/show-86-47.html。↩︎

Once a paper clears the initial screening, the mail review and the desk review, the last two rounds of the Yau Award take place face to face: the regional final (called the "semifinal" in earlier years) and the global final. Both are oral defenses of 30 minutes per team, roughly half presenting and half answering questions. The award rules state that the number of prizes given each year is "determined by a combined assessment of the quality of the entries and the students' defense performance." In other words, by these two rounds a well-written paper only gets you into the room. Your final placing depends on whether you can explain the work clearly and hold up under questioning.

This chapter first sets out the defense rules the organizers have published (Section 1), then covers the differences between the two rounds, building the slides, delivery and timing, and preparing for the judges' questions, and ends with a training timeline and a self-check list. Official rules are cited to their source; experience-based advice is labelled as advice. Treat them differently: the rules must be followed, while the advice can be adapted to your own project.

The Defense Format: Official Rules

This section draws on the following official sources: the FAQ on the Yau Award website (Q12 and Q20, checked 2026-08-31)1, the 2026 rules and schedule2, the 2023–2025 announcements of mainland China semifinalists3, the 2020 and 2021 Finals Student Handbooks4, and the 2024 finals notice5. Details change from year to year; the organizing committee's and each region's official notices for the current season always take precedence.

The Two Defense Rounds at a Glance

The two oral defense rounds compared (mainland China division)
Item Regional final (semifinal) Global final
2026 dates Within the regional review period of 1 October – 2 November; each region announces its own date. Past dates: 28/29 October 2023, 2/3/4 November 2024, 8/9 November 2025 5–6 December at Tsinghua University, with the Science Forum on 6 December
Time structure 30 minutes per team: 15-minute slide presentation + 15-minute Q&A At most 30 minutes per team: 1–2-minute introduction by the advisor + about 13–14-minute student presentation + 15-minute Q&A
Language Official FAQ: no requirement to use Chinese or English before the final (check your region's notice) English only: the final paper, the slides and the defense itself
Who attends Students only; advisors do not attend Students and advisors must both attend; the advisor leaves the room after the introduction
Panel Regional review committee International review committee
Time from shortlist to defense In past years about 9–11 days (2023: list 18 Oct, defense 28 Oct; 2024: list 22 Oct, defense 2 Nov; 2025: list 30 Oct, defense 8 Nov) In past years about four weeks after the finalist list was posted (2025: posted 11 Nov, final 6 Dec)

The last row of Table 1 matters most. The gap between the semifinal list and the defense is usually under two weeks. If you wait for the list before starting on slides and rehearsal, you will be very short of time. Section 6 gives a timeline that starts as soon as the paper is submitted.

Rules in the Final Defense Room

The following rules combine the 2020 and 2021 Finals Student Handbooks and the 2024 finals notice. The three documents agree on the core rules:

  1. The clock starts when you walk in. Timing begins as soon as the team enters the defense room, and the whole session may not exceed 30 minutes. Copying files, opening the slides and setting up equipment all count.

  2. One student may present, or several may share the presentation; any team member may answer questions. The handbooks recommend that every student speaks.

  3. What the defense must cover (official requirement): the background of the problem, the key background material, all of the team's original work, and how the work was divided among team members. This maps directly onto the judging criteria for "academic standards of oral presentation" and "teamwork". Treat it as a checklist your slides must cover.

  4. The advisor's introduction (introduced in 2022): the advisor enters the room, spends 1–2 minutes describing the supervision and confirming the students' identities, then leaves to wait outside. The 2024 notice listed the points to cover: whether the students paid for supervision; whether there are other advisors; whether any training institution was involved; and the students' level of involvement and main contributions. A high-school advisor must also confirm that the students attend their school.

  5. Other people: parents, team leaders and other people may not enter the defense room. (Under the 2024 notice, if the advisor cannot attend for exceptional reasons, one team leader or parent may attend in their place, subject to prior explanation to the committee.)

  6. Waiting and check-in: arrive at the waiting room 30 minutes before your slot; late arrival counts as withdrawal. Bring an ID card or passport (Hong Kong and Macau students bring their travel permit) and a student card. Phones must be silent in the waiting area; you may rehearse quietly without disturbing others.

  7. Equipment: according to the 2024 notice, the room provides a Windows 10 computer, a clicker and other equipment, and slides should be 16:9. Personal laptops are not recommended; if you use one or update your slides on site, the setup time counts toward your 30 minutes.

  8. Submitting materials: after the public posting period, finalists must submit the final English paper and the defense slides through the registration system by the deadline (24:00 on 30 November in 2024). Late submissions are not accepted. If the final paper adds to or changes the posted version, the changes must be listed in detail and uploaded separately under "Other Materials".

  9. Science Gold Award interview: on the morning after the final, the gold-award team from each subject defends again for the Science Gold Award. In past years the list was announced on the evening of the final, so keep your phone on.

Note: the 2021 final was held partly online and partly in person, and the 2022
final was held entirely online for the first time. If this season's defense is
online, follow the current notice on cameras, network, screen sharing and
invigilation, and do a complete test run in advance.

Judging Criteria That Apply Directly to the Defense

Among the subject criteria in Chapter 1, two items are specifically about the defense. The first is academic standards of oral presentation, worded the same way for all six subjects: the defense "must clearly explain the background of the problem, the key background material, and all of the team's original work." The second is teamwork, which is stricter in some subjects than others:

  • Chemistry, Biology, Computer Science: for a team entry, every member must present part of the work orally and answer the judges' questions. The Computer Science criteria also warn explicitly against "free-riding".

  • Mathematics, Physics, Economics and Financial Modeling: every student should "speak on behalf of the team in an organized way".

In addition, the Chemistry criteria require that the research be done by the students under their advisor's guidance, and five subjects (Physics, Chemistry, Biology, Computer Science, Economics and Financial Modeling) state that judges "pay particular attention to the team's academic ethics and integrity". The defense is the judges' most direct way to check this. The committee's official summary of the 2024 final listed the dimensions assessed as innovation, originality, academic standards, teamwork, potential for development, and academic ethics and integrity.

How the Two Rounds Differ, and What to Prepare

The two rounds look alike but emphasize different things, so prepare for them separately.

The regional final decides whether you reach the global final. The judges come from the regional review committee and see every shortlisted team in the region. They have usually read the paper, and their questions tend to target what the paper leaves unclear: whether the method holds up, where the data came from, and which parts the students did themselves. On language, the official FAQ says there is no requirement before the final, but the language of the defense and slides depends on your region's notice. If your paper is in English, ask early and prepare scripts in both languages.

The global final decides the award level. It is run by the international review committee, which includes many scholars based outside China (see the judges' backgrounds in Chapter 4). Compared with the regional final, three things change. First, everything is in English, so the paper and slides must be converted. Second, the advisor attends and describes the supervision, and judges will compare that account with the paper and with what you say. Third, the finalists are closer in quality, so the questioning goes deeper; in mathematics, being asked to derive a key step on the spot is common.

Keep the advisor's introduction, the acknowledgments page and the advisor information form consistent. Many teams overlook this. What the advisor says in the room (whether supervision was paid, whether there were other advisors, whether an institution was involved, what the students contributed) describes the same facts as the acknowledgments page of the paper, the Advisor Information Form submitted at registration, and the Declaration of Academic Integrity. Before the defense, students and advisor should go through these documents together and make sure they agree. Any inconsistency will itself raise questions about academic integrity.

Building the Slides

Time Budget and Slide Count

You have 15 minutes in the semifinal and about 13–14 minutes in the final after the advisor's introduction. At the usual pace of roughly one minute per slide in an academic talk (a rule of thumb), 12–15 main slides is about right. Too many slides force you to rush or skip slides near the end, and the last slides are the results and conclusions, which are the part you should least compress.

Suggested Structure

Table 2 is a suggested structure for a talk of about 13 minutes that suits most subjects. The "Original contribution" and "Team roles" slides correspond to the official requirements above and should not be dropped.

Suggested slide structure (about 13 minutes; for reference only)
Part What to cover Suggested time
Title Title, team members, school; state the research question in one sentence 0.5 min
Background Why the problem is worth studying; what earlier work achieved and what is missing. Only background that bears directly on your work 1.5 min
Original contribution A slide of its own that says plainly "the original work in this study is …", separating background from your own work 1 min
Method Model, derivation, experimental design or algorithm; why this method rather than others 3 min
Results Concrete numbers, charts with error bars, comparison with controls or baselines 3.5 min
Discussion and limitations What the results show; when they do not hold; what remains undone 1.5 min
Conclusion and outlook Restate the contribution; possible next steps 1 min
Team roles and acknowledgments What each member did; the advisor's role; AI tools used, if any (see Chapter 7) 1 min
Backup slides Placed after the final slide and not shown in the talk; used to answer questions you expect Not counted

Details That Matter

  1. Aspect ratio and compatibility. Use 16:9. The room provides a Windows computer, and files made in Keynote or Mac software can show substituted fonts, misplaced equations or broken animations on Windows. Once the slides are final, run through them completely on a Windows machine, and bring a PDF version as a backup.

  2. One point per slide. Put the conclusion in the slide title, for example "Method B reduces error by 38%" rather than just "Results". A judge should know what the slide shows at a glance.

  3. Figures matter more than text. Label axes with units and use font sizes that are readable when projected; show sample sizes and error bars for experimental data; highlight your own method in colour when comparing against baselines.

  4. Number the slides, equations and figures. Judges often say "go back to slide 7" or "your second equation". Numbers save valuable Q&A time.

  5. Keep animation to a minimum. Bullet-by-bullet animations slow you down and often break on a different computer.

  6. Prepare backup slides thoroughly. Put full derivations, extra experiments, ablations, data-source details and parameter tables in backup slides. Pulling one up when a judge asks is far more convincing than describing it.

  7. Stay consistent with the paper. Every number, chart and term in the slides should match the final submitted paper. If you ran extra experiments after the paper was posted, declare the changes under "Other Materials" as required, and say in the defense which results are new.

Delivery and Timing

Dividing the Talk Among the Team

Split the talk by research content rather than by an equal number of slides. A common approach is for whoever led a part of the work to present it and to take the lead on questions about it. In a team of three, for example, one student covers background and the question, one the method, and one the results and discussion. This has two advantages: each speaker knows their part best, and the judges see each member's contribution directly, which is exactly what the teamwork criterion asks for.

Plan the hand-overs in advance with a transition line such as "Now my teammate will present the experimental results." to avoid silences or talking over each other.

Controlling Time

  • Rehearse to one minute under the limit. Aim for under 14 minutes in the semifinal and 12–13 minutes in the final. Walking in, greeting the panel and opening the file are all on the clock, and people usually speak more slowly on the day than in practice.

  • Set time checkpoints. For example, you should have finished the method by the 6-minute mark. Record how long each part takes in rehearsal, and cut content from parts that run over rather than speaking faster.

  • Have a short version ready. If you are interrupted early or run short of time, know which slides can be covered in one sentence and which must be presented in full.

Don't Memorize a Script

A word-for-word script breaks down easily if a judge interrupts or you forget a line, and it sounds recited. Remember two or three points and one key sentence per slide, and say the rest in your own words. A good test: with the slides turned off, can you explain the question, method, results and conclusion of the whole study in three minutes? If you can, you really know the material.

In the final, English is an extra challenge. One practical step is to list the technical terms that recur in your work (method names, metrics, key concepts) with their Chinese equivalents, and say each one aloud until it comes out fluently. Stumbling over your own core terms makes judges doubt how well you know the work.

The Judges' Questions

The Q&A usually decides the judges' overall view of a team. They have read the paper, and they are rarely testing definitions. They want to establish three things: that the work is yours, that you understand the methods you used, and that your conclusions hold.

Common Types of Question

The categories below are compiled from the judging criteria and past participants' published accounts, with an example question for each. For your own project, write at least three specific questions in each category and practise answering them.

Question types and examples
Type Example
Motivation Why is this problem worth studying? What practical or theoretical problem does it address?
Choice of method Why did you choose this approach over a simpler or more common one?
Rigor How do you rule out chance, overfitting or confounding factors? What is your control or baseline?
Mechanism Why does this result occur? Explain the cause rather than describing the effect.
Ownership and roles Which part of this work is your own? What did each member do? What did your advisor help with?
Limits and extensions When does the method not apply? What would you do with another six months?
Literature check Can you summarize reference [X] in one sentence?
Use of AI Which AI tools did you use, and at which stage? Could you have done the work without them? (See Chapter 7.)

What Each Subject's Judges Tend to Probe

Drawing on the analysis of the judges' backgrounds in Chapter 4, follow-up questions in each subject tend to focus on the following (a summary of experience; not every defense will include them):

  • Mathematics: reproducing the proof of a key lemma or theorem on the whiteboard; why a given hypothesis is needed and what happens without it. You must be able to write out the key derivations without notes.

  • Physics: dimensional checks, boundary conditions, the range of validity of approximations; for experimental projects, the main sources of error and how measurements were made. Bringing the apparatus, or photos and video of it, helps.

  • Chemistry: how the product's identity was confirmed (NMR, XRD, mass spectrometry, etc.); why this synthetic route; for computational projects, why this functional and basis set.

  • Biology: where samples and data came from, ethics approval and informed consent; control design, number of replicates, batch effects; in which lab and by whom the experiments were done.

  • Computer Science: how the data were split into training, validation and test sets; whether baseline comparisons are fair; the actual architecture and hyperparameter values and the reasons for them; whether the improvement is significant.

  • Economics and Financial Modeling: the causal identification strategy (DID, IV, RDD, etc.) and whether its key assumptions hold; robustness and placebo tests; whether correlation has been read as causation.

Principles for Answering

  1. Conclusion first, then support. Answer the question in your first sentence, then give one or two sentences of reasoning or evidence. Don't start from the background and circle back.

  2. Use specific numbers and details. "Accuracy on the test set rose from 0.71 to 0.79" beats "it worked better". A specific answer shows you know your own work.

  3. If you don't know, say so, and offer a way forward. Don't make something up, and don't stop at "I don't know." Say that you have not done that part yet and how you would test it, for example "We haven't tested that yet, but we would check it by …". Judges value honesty and the way you think.

  4. Check the question if you didn't catch it. Mishearing in an English defense is common. Restate the question first: "Just to make sure I understand, are you asking about …?" This is far better than answering the wrong question.

  5. Don't pass responsibility to your advisor. "Our teacher told us to do it this way" tells the judges you don't understand your own design. Even if a choice came from your advisor, you should be able to explain the reason for it.

  6. Acknowledge limitations; don't overclaim. If a judge points out a real problem, admitting it and suggesting how to improve earns more trust than arguing. Claim only what your evidence supports.

  7. Agree in advance who answers what. Assign a lead answerer for each area according to your division of work; others may add, but should not jump in or contradict each other. If a judge addresses a question to one person, that person must answer it.

Useful English Phrases

Opening:          Good morning. We are ... from ... Our project is about ...
Transition:       Next, my teammate ... will explain our method.
Checking:         Could you please repeat the question?
                  Just to make sure I understand, are you asking about ...?
Buying time:      That's a good question. Let me think for a moment.
Limitations:      You're right, this is a limitation of our current study.
                  We plan to address it by ...
Work not done:    We haven't tested that yet, but we would check it by ...
Backup slide:     We have a backup slide on this. May I show it?
Closing:          Thank you. We are happy to take your questions.

A Training Plan

Timeline

Because the semifinal usually follows the shortlist by only 9–11 days, start preparing for the defense as soon as you submit the paper on 15 September rather than waiting to be shortlisted. Table 4 uses the 2026 schedule as an example (official notices take precedence).

Suggested defense preparation timeline (2026 schedule)
Period What to do
16 – 30 September (materials review) Finish a first version of the slides and script; build a question list; for maths and physics, practise writing out the key derivations
Early to mid October (awaiting the shortlist) Run at least two full, timed mock defenses; ask teachers or classmates unfamiliar with the project to act as judges; add backup slides for the weaknesses the questions expose
After the semifinal list (about 10 days) Confirm your region's date, venue, language and equipment; run two or three more mock defenses; test the file on the kind of computer used in the room
After the semifinal Write down every question the judges asked and review which answers were weak; start converting the paper and slides to English
Finalist posting period (early to mid November) Finish the English paper and slides; run mock defenses in English; check with the advisor that the introduction matches the acknowledgments page and the information form
Before the submission deadline (end of November in past years) Submit the final English paper and slides on time; if the paper differs from the posted version, upload a list of changes under "Other Materials"
The week before the final Give the full talk at least once a day; prepare IDs and a USB drive (slides and PDF); confirm travel and arrival time

Running a Mock Defense

  1. Follow the real format exactly. Start the clock when you knock and walk in; 15 minutes presenting, 15 minutes of questions, no pauses. For the final, have the advisor give the 1–2-minute introduction as well.

  2. The judges should not be your advisor. Your advisor knows the project too well to ask an outsider's questions. Ask other teachers in the field, or university students, and include at least one person from a different field.

  3. Record and review. Check three things: did you run over; did you read from the screen; did the first sentence of each answer actually answer the question?

  4. Build a question bank. Record every question from each mock and the key points of your answer, and practise the weak ones repeatedly. Keep the real semifinal questions in particular: the final often asks the same kinds of question, only in more depth.

Defense Self-Check List

Go through each item before the defense. Anything you cannot tick needs fixing now:

  1. We have confirmed the date, venue, language, equipment and submission requirements for this round from the current official notice.

  2. Our slides are 16:9, have been played through completely on a Windows computer, and have a PDF backup.

  3. The slides include a separate slide distinguishing background from original work, and a slide on each member's role.

  4. Every member presents a part and can answer questions on their own part independently.

  5. The full talk fits within the time limit with at least one minute to spare.

  6. We can explain the question, method, results and conclusions without the slides.

  7. Every number and chart in the slides and talk matches the final submitted paper.

  8. We can state the main finding of every major reference we cite in one sentence.

  9. (Maths, physics, etc.) We can write out the key derivations on a whiteboard without notes.

  10. We have backup slides covering the common questions, and we know where each one is.

  11. If we used AI tools, we can say which ones, at which stage, and which work we did ourselves, and this matches the acknowledgments page.

  12. The advisor's introduction matches the acknowledgments page, the Advisor Information Form and the Declaration of Academic Integrity.

  13. (Final) The English paper and slides were submitted before the deadline, and any changes from the posted version have been listed and uploaded.

  14. IDs, student cards and the USB drive are ready, and we plan to reach the waiting room at least 30 minutes early.

A closing word. The 30 minutes of the defense are the judges' only chance to meet you and your research in person. They do not expect a high-school talk to be as polished as a conference presentation. What they want to confirm is that the research really is yours, that you really understand it, and that you can say honestly what it is worth and where it falls short. Getting those three things across matters more than beautiful slides or fluent English.


  1. https://www.yau-awards.com/page-assessment.html↩︎

  2. https://www.yau-awards.com/page-rule.html; https://www.yau-awards.com/show-86-57.html, published 2026-04-16.↩︎

  3. https://www.yau-awards.com/show-86-34.html, https://www.yau-awards.com/show-86-42.html, https://www.yau-awards.com/show-86-50.html.↩︎

  4. Official download page: https://www.yau-awards.com/lists-download.html.↩︎

  5. An English summary of the 2024 finals notice is at https://en.embarkchina.org/blog/1784.html; the committee's official summary is at https://www.yau-awards.com/show-86-47.html.↩︎