The contrast is between two hiring signals, not two levels of difficulty

One October 2026 NAVER vacancy is explicitly an internship, but it requires at least one authored publication, or an accepted forthcoming publication, at a top-tier AI conference or journal. Another is a full-time robotics-engineering opening at NAVER LABS. It includes a coding exercise, yet the posting says the exercise is for the first interview and does not itself determine pass or fail.

The meaningful difference is not that interns have to code and full-time engineers do not. Both roles demand technical expertise; the two postings simply describe different ways of showing it.

The internship targets original world-model research

NAVER Future AI Center's Generation Research team advertised World Models for Physical AI Research on October 19. Applicants may hold a related bachelor's degree or be about to graduate, but the authored top-tier paper is mandatory. Team research covers controllable world models conditioned on action signals, use of real-world data, generalization to new environments and generation grounded in measured scenes.

These are research questions where a candidate's ability to formulate, test and communicate an idea matters. But the publication requirement does not make a coding assessment irrelevant: the advertised selection process still includes document screening, a coding test and a role interview. Its exact grading weight is not disclosed.

The full-time opening reaches from demonstrations to robot deployment

Published on October 22, NAVER LABS' Robot Manipulation Data Engineer role covers collecting and curating teleoperation and demonstration data for vision-language-action (VLA) training. Duties also include synthetic or simulated data, pipeline quality, model training and evaluation, and deployment on actual robots.

The advert allows new or experienced applicants as a category, while separately requiring a relevant master's or doctoral degree OR at least five years of research/development experience. It also specifies imitation learning, robotics control, large-scale datasets and programming experience. That distinction matters: 'open to new graduates' does not cancel the academic or professional threshold.

The coding test is still administered

The stated sequence is document screening, telephone interview, first interview with a preliminary coding test, second interview, and final hiring steps. NAVER LABS specifically says the coding test is used as reference material during the first interview and is not a separate pass-fail assessment.

It is therefore incorrect to say the company eliminated coding tests, or that programming skills are no longer assessed. Nor does this single advertisement establish a uniform selection policy for all NAVER LABS roles.

Research discovery and robotics implementation have different bottlenecks

A controllable world-model researcher investigates how an AI system represents changes in an environment and generalizes to new contexts. A robot-manipulation data engineer must build the evidence and infrastructure to collect usable data, train and evaluate models, and deploy them on machines. The activities interact, but the talent question is different.

Employers risk blurring their hiring criteria by treating both as one generic 'AI robotics engineer' role. It may be more useful to define whether success means advancing a research hypothesis or producing a repeatable operational pipeline before deciding how to interview.

What applicants should actually prepare

For the research internship, first verify the publication eligibility requirement. Be ready to explain the problem addressed, the experimental design, the results, and your individual contribution. Do not assume bachelor's-level eligibility means entry-level research expectations.

For the NAVER LABS role, the absence of a standalone coding-test rejection is not an absence of technical scrutiny. Evidence of building datasets, solving simulation-to-robot integration problems, and explaining deployment decisions is likely to be more informative than listing libraries alone. These are editorial preparation suggestions, not the company's published scoring rubric.

Separate deadlines and separate organizations

The NAVER Future AI Center internship closes on October 27, 2026 at 6 p.m. Korea time. NAVER LABS lists November 1 at 11:59 p.m. Both notices allow changes or earlier closure under stated conditions. Applicants should check the official notices before applying.

The two postings do not prove a coordinated robotics project, an overhaul of NAVER's overall hiring system, or the same role-level standards across subsidiaries. They establish one narrower lesson: technical hiring needs evidence matched to the actual work.

BANSEOG VIEW

BANSEOG VIEW | Define the problem before choosing the hiring test

Employment labels are poor substitutes for the actual capability a team needs. Here the internship asks for documented research output, while the full-time robotics job specifies hands-on data and deployment work and contextualizes its coding exercise.

Original research and operational robotics engineering should not automatically be assessed with an identical screen. Employers can begin by identifying the work product and failure mode; applicants can begin by identifying which problem their past evidence proves they can solve.

This interpretation is confined to two published vacancies, not a claimed company-wide shift in NAVER's assessment policy.

SOURCES

Primary sources and references

Official NAVER Careers listings checked October 25, 2026. The research internship's top-tier authored paper is required; a robotics conference publication is preferred, not required, for the NAVER LABS role. The NAVER LABS role lists the career category as open but separately imposes a master's/PhD or five years' R&D threshold. Its coding test is conducted but is not a standalone pass-fail gate; the intern test's weighting is not disclosed. Do not infer a company-wide policy change or a shared project.