打造最強教學 AI
重新定義未來的學習
Build the Ultimate AI Tutor
Redefining Future Learning
第一屆教學怪物挑戰賽:
尋找能深刻理解學習需求、生成適性化影音教材的頂尖 AI Agent。
The 1st Teaching Monster Challenge: Seeking the top AI Agents capable of understanding learning needs and generating adaptive educational videos.
競賽宗旨 Mission Statement
「生成影片」不等於「有效教學」
我們已經見證了 AI 根據 Prompt 生成高品質影片的能力。
然而真正的教學不僅要求內容的正確性,更需要具備 學科教學知識 (PCK) 的智慧:它不僅要懂知識,更要懂「如何教」。
"Generating video" is not "Effective Teaching"
We have already witnessed AI's ability to generate high-quality videos based on prompts.
However, true teaching requires not only the accuracy of the content, but also the wisdom of Pedagogical Content Knowledge (PCK): it requires not only understanding the knowledge, but also understanding "how to teach".
挑戰重點
如何將複雜的概念拆解、建立邏輯連結、針對不同背景進行差異化教學,並維持學習者的專注力。
Key Challenges
How to break down complex concepts, establish logical connections, provide differentiated instruction for different backgrounds, and maintain learners' focus.
競賽背景與範疇 Scope & Description
-
01
核心任務Core Task
開發一組 AI Agent,透過 API 接收文字學習需求,全自動生成教學影片。 Develop an AI Agent that receives text-based learning needs via API and fully automatically generates teaching videos.
-
02
學科領域Domain
中等教育階段之 物理、生物、電腦科學、數學 Physics, Biology, Computer Science, and Mathematics in Secondary Education.
-
03
語言要求Language
生成影片以 英文 (English) 為主。 The generated videos are primarily in English.
-
04
呈現形式Format
時長上限 30 分鐘。
風格不拘(動態投影片、AI 數位人、動畫圖表皆可)。 Maximum Video Duration: 30 minutes. (Decided by the Agent).
Style: Open (Slides + Voiceover, Digital Avatar, Motion Graphics, etc.).
挑戰任務與規則 Tasks & Rules
⚙️參賽技術規格 ⚙️Technical Specifications
API Input (JSON)API Input (JSON)
{
"request_id": "1314520",
"course_requirement": "Self-Attention Mechanism",
"student_persona": "High schooler, no calculus."
}
API Output (JSON)API Output (JSON)
{
"video_url": "https://yourapi.com/video.mp4",
"subtitle_url": "https://yourapi.com/subtitle.vtt",
"supplementary_url": ["https://yourapi.com/slides.pptx","https://yourapi.com/document.pdf"]
}
除 "video_url" 外皆為可選,也就是可以僅包含 "video_url",且"supplementary_url"可接收 null、字串、字串清單(含空清單),最多 5 個檔案,合計不超過 100MB,詳細資訊請參考規則All fields except "video_url" are optional, meaning you can include only "video_url." Furthermore, "supplementary_url" can accept null, strings, and lists of strings (including empty lists), up to 5 files, with a total size not exceeding 100MB. Please refer to the rules for more details.
系統限制Constraints
- 回應時限:30 分鐘內回傳。Latency: Within 30 mins.
- 全自動化:禁止人工介入。Autonomy: No human-in-loop.
- 工具使用:允許聯網、Interpreter。Tools: Search & Code Interpreter OK.
公開基線模型管線Public baseline model pipeline
可以參考官方的 GitHub 儲存庫,包含從接收請求到生成影片回傳的所有流程You can refer to the official GitHub repository, which contains the entire process from receiving a request to generating and uploading the video.
📝評分標準 📝Evaluation Criteria
初賽由「AI 學生」與「人類評審」評比,決賽由教育專業評審評比。
The preliminary round is evaluated by AI Students and human raters; the final round is evaluated by education professionals.
引用規範 Zero Hallucination
Proper Citations
敘事連貫 Scaffolding
Coherence
差異化教學 Student Level Awareness
Differentiated Teaching
生動有趣 Multimodal
Vivid & Engaging
賽制流程 Competition Phases
所有時間皆以 Anywhere on Earth (AoE, UTC−12) 為準
Competition Phases All times are in Anywhere on Earth (AoE, UTC−12).
🚀 熱身賽第一輪(32 題) 🚀 Warm-up Round 1 (32 questions)
釋出熱身賽題目,由「AI 學生」提供自動化評測報告,協助團隊優化模型,「AI 學生」的評分僅供參考,不影響比賽成績。
The warm-up competition questions have been released, and "AI Student" will provide automated evaluation reports to help teams optimize their models. The "AI Student's" scores are for reference only and will not affect the competition results.
🔍 熱身賽第二輪(32 題) 🔍 Warm-up Round 2 (32 questions)
釋出共 32 題更貼近初賽難度的題目,由 AI 學生評測,供隊伍進行最後調校。
Release 32 topics closer to preliminary difficulty for final model adjustments.
⚔️ 初賽第一階段(16 題) ⚔️ Preliminary Phase 1 (16 questions)
釋出初賽第一階段題目。
The questions for the first stage of the preliminary round will be released.
⚡ 初賽第二階段(16 題) ⚡ Preliminary Phase 2 (16 questions)
釋出第二部分初賽題目 16 題,5/15 00:00:00 AoE 截止(僅兩天)
第二部分題目模擬決賽情境,僅給予極短的作答時間,測試系統在時間壓力下的表現。初賽成績由兩部分題目的成績共同計算。
Release the second part of the preliminary round, consisting of 16 problems. The deadline for AoE submissions was May 15th at 00:00:00 (only two days).
The second part of the questions simulates the final round scenario, giving the system only a very short time to answer, testing its performance under time pressure. The preliminary round score is calculated by combining the scores from both parts.
🛑 初賽截止 🛑 Preliminary Deadline
兩階段皆於 5/15 00:00:00 AoE 截止。AI 學生依四大評分標準進行自動化評分,選出最多 10 名進入真人評審階段。入選隊伍的影片交由透過 Prolific 平台招募的真人評測者進行盲測評比。評測者會隨機看到兩個 AI Agent 針對同一題目產生的影片,判斷何者較佳。依據 Elo 積分排名,選出前三名進入決賽。
Both phases will close at 00:00:00 on May 15th (AoE). AI students will automatically score the videos according to four evaluation criteria, selecting a maximum of 10 teams to advance to the human review phase. The selected teams' videos will then be blind-tested by human evaluators recruited through the Prolific platform. Evaluators will randomly see two videos generated by the AI Agents on the same topic and determine which is better. Based on Elo ratings, the top three teams will advance to the finals.
📢 決賽名單公告 📢 Finalists Announcement
公布決賽三強名單。
Announce top 3 finalists.
🏆 決賽巔峰 🏆 Grand Final
由教學專家擔任評審並出題,共 16 題。每位評審各自對一部分影片排 1-3 名,以所有排名的平均值決定最終名次。每題生成時限 30 分鐘,測試次數無上限,可重傳(以最後一次提交為準)。
16 questions will be set by teaching experts. Each judge independently ranks a subset of videos from 1st to 3rd. The final standing is determined by the average of all rankings. Each question has a 30-minute time limit for generation, and there is no limit to the number of tests. Questions can be resubmitted (based on the last submission).
🤝 線上成果交流會 🤝 Online Workshop
技術交流、公布名次、頒獎(6/25 18:00-20:00 AoE, 6/26 14:00-16:00 UTC+8)。
Technical exchange, ranking announcement, and awards (June 25 18:00-20:00 AoE, June 26 14:00-16:00 UTC+8).
評審陣容 Meet Our Judges
台灣師大科教中心 NTNU Science Education Center
台灣師大物理系 Department of Physics, NTNU
台灣師大生命科學專業學院 School of Life Science, NTNU
台灣師大科學教育研究所 Graduate Institute of Science Education, NTNU
三民國中 Sanmin Junior High School
北一女中 Taipei First Girls' High School
台大師資培育中心 Center for Teacher Education, NTU
台大師資培育中心 Center for Teacher Education, NTU
台灣大學人工智慧頂尖研究中心 AI-CoRE, NTU
台灣大學人工智慧頂尖研究中心 AI-CoRE, NTU
台灣大學人工智慧頂尖研究中心 AI-CoRE, NTU
關於比賽名稱:「教學怪物」的由來 About the Name: Origin of "Teaching Monster"
你可能會好奇,為什麼一場結合教育與 AI 科技的創新賽事,會以「怪物」來命名?這個名稱背後,其實蘊含了對兩大流行文化的致敬,以及我們對未來教育的深切期盼。
You might wonder why an innovative competition combining education and AI is named "Monster"? Behind this name lies a tribute to two major pop cultures and our deep expectations for the future of education.
源起一:現象級實境秀《中文怪物》 Source 1: The Reality Show "Chinese Monster"
前陣子,由百萬 YouTuber「酷的夢」斥資打造的大型實境節目《中文怪物》席捲社群,引發熱烈迴響,掀起了以「XX 怪物」為競技命名的風潮。在這裡,「怪物」一詞跳脫了傳統的負面意涵,轉而象徵在特定領域中實力頂尖、打破常規的卓越佼佼者。我們希望透過這場比賽,讓 AI Agent 成為能顛覆傳統教學框架的「教學怪物」!
Recently, the large-scale reality show "Chinese Monster," created by million-subscriber YouTuber "Ku's Dream," swept through social media and sparked a trend of naming competitions "XX Monster." Here, the word "Monster" transcends its traditional negative meaning, symbolizing top-tier, rule-breaking excellence in a specific field. We hope that through this competition, AI Agents will become "Teaching Monsters" that subvert traditional educational frameworks!
源起二:動漫《暗殺教室》的極致「因材施教」 Source 2: Individualized Instruction in "Assassination Classroom"
名稱的第二個靈感,來自日本經典動漫《暗殺教室》。劇中的「殺老師」是個「怪物」,擁有 20 馬赫的移動速度,能在同一時間內,為班上每一位學生量身打造專屬教材,並進行一對一精稱指導。這種「全員客製化學習」的極致境界,過去或許只存在於想像中;但在今天,AI Agent 有可能可以針對不同學生的需求與弱點,即時生成並調整最適合的學習內容,成為每個人的「殺老師」。
The second inspiration comes from the classic Japanese anime "Assassination Classroom." The character "Koro-sensei" is a "monster" with a speed of Mach 20, capable of creating personalized materials and providing one-on-one guidance for every student in the class simultaneously. This ultimate level of "fully customized learning" might have only existed in imagination before; but today, AI Agents can potentially generate and adjust the most suitable learning content in real-time based on each student's needs and weaknesses, becoming everyone's "Koro-sensei."