Translated by Technical Blog Translator
Written Later
Updated May 17, 2026
Scanned through after reposting, and I feel a bit regretful—I should have written about the experiences along the way (mainly wanted to post photos of Shimokitazawa). I’ll dig a hole here for now, and maybe write a separate post when I have time.
Also, here are some application results for reference. Looking back, the application season wasn’t too bad; at least I got what I wanted. Whether I’ll actually go is a matter for the second half of the year. (The submission time was roughly after November 20, 2025; if the application opened after that, I submitted as soon as it opened.)
- HKU CS, con offer 80, declined
- UCL CGVI con offer 85, declined
- TCD VR con offer 80, declined (by the way, this one came super fast, probably because no one applies)
- Edinburgh CS con offer 80, declined, waitlist can extend to June… impressive
- IC Advanced Computing, rejected
- KTH CS, rejected
- Aalto Game Design, rejected (this program is probably really difficult)
- NUS Computing, rejected
Only for UCL CGVI did I seriously write a statement of purpose; the rest used a generic template casually, which probably had a big impact. Also, my publication news wasn’t out when applying, which also had some effect. Other than IC where I probably didn’t meet the GPA cutoff, for the rest I should have met the bar.
Moreover, surprisingly, Sino-US relations have eased recently, so maybe I can go to the US next year.
The following is the original post content
I participated in the AY2026 University of Tokyo Creative Informatics entrance exam (winter entry, February 2026) and passed, so I’m sharing my experience as usual.
Foremost
The University of Tokyo’s Information Science and Technology application process has two steps. First, you need to write a Research Proposal (documents) for the first round of screening, which filters out about 30% of applicants (mainly depending on how many apply, as the number of exam takers each year is basically fixed; you can check past document screening results on the Internet Archive for reference). After passing the document screening, you need to take the entrance exam, which usually consists of two tests; different specializations have different papers. Only after passing the entrance exam can you enroll.
Also, contrary to common belief, UTokyo IST applications do not require any Japanese language proficiency, but you do need a TOEFL score of roughly 90 or above.
My Background
Undergraduate institution: Beihang University
Major: Computer Science and Technology
GPA: 87/100
TOEFL: 105
One internship and one SCI Zone 2 paper, though these experiences didn’t matter much—it mainly depends on the exam score.
Timeline
UTokyo’s master’s programs usually have summer and winter entries. Winter entry is essentially a supplementary intake for summer, so there are relatively fewer specializations recruiting (for example, CS only recruits in summer).
Normally, the specific details for the next year’s admissions are released each March at https://www.i.u-tokyo.ac.jp/edu/entra/entra_e.shtml, so from that time on, you need to pay attention and prepare your documents. The summer document screening is around May, and winter around November. Before that, it is currently recommended to contact the professor by email in advance (it’s advised to have an RP ready). Once you pass the document screening, just take the exam on schedule. Overall, it’s recommended to prepare at least half a year in advance. The specific timeline is as follows:
In March, the admissions website updates; understand roughly what’s needed, and have your TOEFL score ready.
June: Start preparing the RP. Since I applied to Professor Takeo Igarashi’s lab (he explicitly stated on the lab’s recruitment page that early contact is required), I put a lot of effort into the RP, spending over a month.
July: Started reviewing basic knowledge while interning (skimmed through a book). After the internship ended in September, I began working on problems, going through all previous Creative Informatics exam questions in order.
The following year (this year): At the end of January, traveled to Tokyo on a tourist visa for the exam.
Documents
Seeing that many people were rejected at the document stage this year, I want to remind future applicants to pay more attention to the documents; first impressions are quite important.
To prepare the RP, first look at the recent work of the lab you want to apply to, then find a few that interest you and read them carefully. After that, look at the main conferences in the field’s recent publications. If something catches your interest, look at it a bit more closely. In any case, your RP should have at least either creativity or a trendy research direction. Once you have a rough idea of “what the problem is” — “some related work” — “how you plan to solve it,” you can start writing. Basically, clearly write the problem, related work, what you want to do (you could also add what limitations your method might have), and a timeline for how you plan to address the problem during your master’s.
I thought Professor Igarashi would be strict about this, so I put a lot of thought into it and wrote an RP of over 10 pages to send. Fortunately, less than half an hour after contacting him, the professor replied “come take the exam” (really fast), which was a relief.
Exam Preparation
For the exam, you definitely need to practice past papers. I think you should at least do all the questions from the past 4 years, and you can refer to The Kai Project. Older past papers can be found on the Internet Archive, they’re all there.
For Creative Informatics, the exam scope is somewhat vague. Based on the questions in recent years, the focus is on probability theory (especially stochastic processes, which will be covered later), algorithms, and networking. Of course, other topics that have appeared also need to be reviewed. Overall, refer to the following books:
Mathematics (calculus, linear algebra, probability theory): I only looked at the linear algebra book (“Introduction to Linear Algebra”), and also watched Andrew Ng’s math tutorials. Pay special attention to stochastic processes.
Computer Architecture and Digital Circuits (digital circuits are frequently tested): “Digital Design and Computer Architecture, Second Edition – David Money Harris, Sarah L. Harris”
Operating Systems: “Operating System Concepts (Abraham Silberschatz, Peter B. Galvin etc.)”. This is rarely tested, so you can skip it if time is tight; this book is just average.
Compilers, Automata Theory (automata theory is key): Skimmed the Dragon Book for compilers, and for automata theory read “Introduction to Automata Theory, Languages, and Computation ( etc.)”
Algorithms (key): Of course, “Introduction to Algorithms (Thomas H. Cormen, Charles E. Leiserson etc.)”
Computer Networks (key): “Computer Networking A Top-Down Approach (James F. Kurose, Keith W. Ross)”
Terminology Explanation
The last question of the Creative Informatics exam is terminology explanation, which is the final mercy of this paper. Naturally, you should aim for near-perfect scores. To prepare for this part, I made a small webpage. After organizing it, I will open-source it here for your ease of use.
Programming
The practical exam for Creative Informatics is programming. You are allowed to use your own computer for the exam and can use offline code. However, you cannot connect to the internet or use generative AI during the exam.
I have a solid programming foundation, so I just went through the questions once and prepared libraries and algorithms that might be used. In recent years, the questions basically involve implementing an algorithm they describe, and then at the end giving you a few large data points to compute.
If you don’t have much programming experience, you definitely need to practice LeetCode’s Hot 100.
Exam Retrospective
The first exam was theory. The first big question gave a self-defined floating-point binary format and asked you to compute various things, plus some proof questions. It was basically a reading comprehension; the calculations were fairly simple, but I didn’t quite understand the last proof question, so completion was about 90%.
The second big question was rather disgusting—it gave a problem about stationary processes. Coincidentally, my probability theory was weak (and I hadn’t systematically reviewed stochastic processes before the exam, only looked at Markov process calculations). Surprisingly, I managed to derive it step by step, completing 100%. To be honest, that day was probably the best my stochastic processes ever was in my life.
Finally, for the terminology explanation, I chose four that I was confident in, answered with diagrams, and completed 100%.
The second session’s programming exam had a coordinate system area calculation. Essentially, it gave a bunch of points and you used the given algorithm to compute the result. There were quite a few questions, but implementing for small data sizes was quite straightforward.
However, the last two questions with large data sets were problematic. I spent over an hour but still couldn’t come up with a significantly optimized algorithm (multithreading was a bit tricky to write at the time, and my MacBook Air’s performance was pitiful). Overall completion was about 80%.
Interview
The interview was completely different from what other experience posts said—it wasn’t relaxed at all. I had the flu that day and was a bit dizzy. At the start, they said I could introduce myself using two slides (which wasn’t mentioned beforehand), and I was momentarily stunned. Then, for the RP, I was supposed to use diagrams to explain, but I forgot to set up Zoom permissions in advance. I was chased with questions, and honestly, my answers were quite mediocre (I was full of regret afterward).
But the interview is probably just a formality; as long as you pass the preliminary contact, it shouldn’t be a big issue (contacting professors early is really important).
Final Words
Recently, relations between the two countries aren’t great, and to be honest, I’m not very optimistic about Japan’s future development prospects, so I’m still considering whether to go (I simultaneously applied to Hong Kong and Europe). Overall, competition at UTokyo has been getting fiercer year by year. The acceptance rate for the Creative Informatics entrance exam is about 30% (counting the document stage, it’s only about 15%). Since studying in Japan doesn’t rely heavily on GPA and is cheap, UTokyo is still generally recommended (and personally, I have a pretty good impression of Tokyo).
Finally, here’s a doujinshi I finally managed to buy:
