mastAdventCalendar2025

[CAUTION]
Note: This article was translated from Japanese through Plamo’s translation system. This is purely for fun.

This is the eighth entry for the “mast Advent Calendar 2025”.
Day seven was written by MY with the article titled [title omitted].

This article is authored by klis third-year student Romanov.
Hello to all mast members—this is my first time writing for mast.

Motivation Behind Writing mastAdC

At klis, we conducted a laboratory assignment process—some might call it a “war”—throughout October. As a result, I was assigned to the Machine Learning and Language Understanding Laboratory (hereafter: wkb Lab). Wkb Lab belongs to mast, not klis. While you can check their official website to see what kind of research they do, I’ve included the following excerpt as a reference:

Our research fundamentally seeks to answer the question: "What does it mean to understand language?" To address this, we employ machine learning approaches. Furthermore, we examine how insights gained through this research endeavor can contribute to the development of artificial intelligence (AI) technologies for language processing.

It’s not particularly unusual for klis students to be assigned to mast labs—there are several other cases like the otiAI Lab and M&I Lab, where mast labs accept applications from klis students as well (and if you count unofficial routes, the number would be virtually limitless).
That said, the sheer number of such cases remains relatively small, which is why I wanted to document the process leading up to my assignment to the wkb Lab—this was the motivation behind writing this Advent Calendar entry.

My Encounter with wkb Lab: The World of Intellectual Pursuit I-1

At klis, we offer a semester-long course titled “The World of Intellectual Pursuit I-1.” This course is taught by wkb-sensei, who serves as supervisor for the wkb Lab. While the syllabus contains detailed information, in broad terms, it can be summarized as follows:

An Advanced Research Experience Program (ARE) with Stronger Lecture-Based Components

ARE is a support framework designed to assist motivated first- through third-year students in research activities. Upon selection, participants can conduct research under a professor (their advisor) even before being assigned to a laboratory. While “The World of Intellectual Pursuit I-1” does have the constraint that the theme must be within the professor’s specialized field, it allows students to engage in year-long research under a professor’s guidance—similar to ARE. However, it requires less autonomy than ARE; instead, the research and course direction are determined through discussions with both the professor and other participants. Therefore, specific content likely varies from year to year.
Since I am currently enrolled in this course this year, though we’re still in the intermediate stages, here are the main activities we’ve undertaken: (I’m keeping this brief as detailed explanations would deviate from the main point.)

• Reading academic papers (specifically papers on multi-agent reinforcement learning and emergent communication fields)
• Conducting research experiments using the provided experimental code
• Participating in regular discussions with the professor and other students in the laboratory

At the start of the current academic year, I lacked both the knowledge and implementation skills to independently pursue research or experiments, so my progress has largely been guided by Professor Wkb. However, I feel I’ve gained some knowledge recently. (My implementation skills remain somewhat limited, though.)
It’s unclear yet whether I’ll be able to produce results at the level of research presentations… (I’ll do my best, of course.)

I realize I’ve veered somewhat off-topic, but through this course, I first became aware of Professor Wkb’s research group. Additionally, I’ve gained valuable insights into the lab’s atmosphere and Professor Wkb’s philosophical approach, so I’m glad I took this course.

I can’t remember the exact course name, but there’s another course with the same content under a different title available on MAST.

Summer Break

Not long after learning about Wkb’s research group, I also discovered that klis offered placement opportunities there, though it wasn’t particularly high on my list of preferred options at the time. However, I did express interest in joining a machine learning-related lab, so I wasn’t completely unaware of the possibility. Apologies for the vagueness…

Returning to my original point, I’ve always aspired to work in game development, and during summer break, I applied to and participated in several internships in the field. During these experiences, I frequently heard discussions about how Reinforcement Learning (RL) was gaining significant attention. (To put it broadly, this has been the case for some time now.)
Since Wkb’s research extensively incorporates RL-related technologies, I considered that there might be some future alignment between my interests and their research focus.

This gradually built up my inclination toward joining Wkb’s lab.

Around the beginning of summer break, I came across a blog written by an alumni from Wkb’s lab who now works as a researcher. This discovery somehow made the prospect of joining Wkb feel less daunting.

Lab Placement

As mentioned in the preceding section, by summer break I didn’t have a clear “I absolutely must join ○○ Lab!!” type of preference, so when I heard in October that lab placement interviews were becoming available, I became quite anxious.
Nonetheless, by the end of the first week of October, I had decided which labs to visit for interviews and made the necessary appointments. From the beginning, Wkb Lab and YM Lab both aligned well with my research interests and offered significant benefits, so I narrowed my consideration to these two options. In hindsight, choosing these two labs turned out to be quite a wise decision - I heard that ML-related labs like M&I Lab and KaTo Lab had extremely competitive applicant pools this year. In fact, I couldn’t help but wonder why there were fewer applicants for attractive labs like Wkb and YM Lab. (This isn’t meant in any negative way.)

Since I was taking “The World of Intellectual Exploration,” I had been maintaining regular contact with Professor Wkb, but actually expressing my lab preference was when I visited the lab the following week for an unrelated matter. During the interview, we focused less on what specific research I wanted to do and more on verifying whether my understanding of the Wkb Lab’s specialized field was accurate. I also mentioned that I was particularly drawn to the lab’s use of RL as a methodology. It was reassuring that the atmosphere suggested I could decide what specific research to pursue and what kind of research I wanted to do after being assigned to the lab.

As for negative considerations, I explicitly stated I didn’t want to work on LLM prompt engineering research, but since LLMs themselves were largely outside the lab’s scope, this wasn’t a problem. (This was just my personal preference - there are many researchers doing fascinating and meaningful work in this area.)After a similar interview at YM Lab and receiving positive feedback, I hesitated but ultimately selected the Wkb Lab as my first choice when registering my preferences.

“Hope” refers to the system used by klis students to register their lab preferences and check the number of applicants for each lab.

As a result, I received a job offer around mid-October and was officially assigned to the Wkb Lab.

What was this all about?

After writing all this, I’ve realized I may have lost track of what the actual topic was.
When put into words, it turns out to be quite straightforward: I became familiar with the professor through lectures, our interests aligned, and that’s why I decided to join their lab. However, this seemingly ordinary process might actually be more difficult than it seems, and not many people get to experience it.

Reflecting on my first and second years as an undergraduate, I rarely had opportunities to sit face-to-face with professors and discuss my academic interests—or even develop the kind of relationship where such conversations could naturally occur. Even when such opportunities presented themselves, I lacked the courage to pursue them. Having interests but lacking knowledge often leads to one-sided discussions or debates, making it difficult to cultivate any further relationship afterward. Surprisingly, this path might actually be quite challenging.

Yet when third-year students reach the lab assignment period, suddenly matching with professors and research labs becomes the single most important consideration from the student’s perspective. Without any prior interaction, judging this match based solely on information from websites and course syllabi is like entering battle without proper equipment—resulting in what amounts to an “unproductive information war” where students submit lab preferences without truly knowing what they’re getting themselves into. (There are also limitations to what can be discussed during interviews.)
In particular, requesting a lab outside your own department without any prior preparation may avoid internal departmental competition, but it often leads to even greater information gaps.

The solution to information shortages is obvious: start moving sooner rather than later. However, few people actually take immediate action. Even my decision to take certain courses was prompted by a friend’s invitation…

For this reason, I hope this article can at least partially fill in some gaps by sharing one sample case of joining a lab outside your own department.

Since many programs seem to use GPA-based “rock-paper-scissors” for placement, this information may not be particularly useful for actual placement decisions.

That concludes Day 8 of the mastAdC2025 series. To all mast-affiliated students, should we cross paths in some way, please do make an effort to get along!