Insiders expose flaws in Korea's college entrance exam

An examinee prepares for the College Scholastic Ability Test (CSAT) for the 2026 academic year at Kyungbock High School in Jongno-gu, Seoul, on Nov. 13, 2025. Korea Times file

There is a test that is simultaneously praised as "the world's most perfect multiple-choice exam" and criticized as "the main culprit for ruining classroom instruction." It is the College Scholastic Ability Test (CSAT), known locally as Suneung.

Launched in 1993 to move Korean education away from rote learning, the CSAT is now facing growing calls for a major overhaul in the era of artificial intelligence.

Critics argue that the 33-year-old exam has completely lost its way after a series of patchwork fixes aimed at addressing excessive academic pressure, entrenched university hierarchies and soaring private education costs.

The Hankook Ilbo conducted an in-depth investigation into the crisis triggered by last year’s controversial “Bul-suneung,” or “fire test,” which drew widespread criticism for its extreme difficulty. Based on interviews with twelve key insiders — including former institute presidents and CSAT question writers — the report examines how the long-standing exam must evolve.

"The original plan was to hold the exam six times a year. If a student scored over 250 points (out of 400) just once, they wouldn't need to take it again. We believed that score alone was enough to prove the ability to study at a university."

Park Do-soon, a professor emeritus at Korea University and the first president of the Korea Institute for Curriculum and Evaluation (KICE), is considered the principal architect of the CSAT. He originally envisioned the test as a simple qualification exam, with a passing score equivalent to today’s Grade four or five level. The reality, however, has drifted far from that blueprint.

Since its inception, the exam’s influence has grown overwhelmingly. Because the status attached to the university name — a key factor in Korea’s social and economic hierarchy — depends heavily on CSAT scores, students and their families became extremely sensitive to even a one-point difference. As a result, nine of the thirteen former KICE presidents were unable to complete their terms. Oh Seung-geol, who resigned last December, was the first to step down not because of a mistake in the test, but simply because the exam was deemed “too difficult.”

A sorting tool for top medical schools in Seoul

“To rank students, super-high difficulty questions are necessary. Even if people criticize them, those questions are doing their job.”

Jeong Moon-seong, a professor of social education at Gyeongin National University of Education and the 2024 CSAT committee chair, defended the controversial “killer questions” — questions with a correct-answer rate of 20 percent or lower, often called “monster questions." In a five-option multiple-choice test, a 20 percent accuracy rate is statistically equivalent to nearly every student simply guessing.

However, question writers unanimously argue that "killer questions shouldn’t be blamed." Under the current admissions system, the CSAT's primary role is to sort students by score so universities can select them efficiently.

This, they say, is the inevitable outcome of the test shifting away from its original purpose — assessing basic academic qualification — and becoming a tool to sort students by score.

A teacher involved in writing questions for the Korean-language section said exam creators fear an easy exam far more than a difficult one.

“For the exam committee, a ‘Mul-suneung’ — an overly easy test — is far more frightening than a difficult test. If the test is too difficult, students simply complain. But if it’s too easy, it creates chaos. It means the CSAT failed in its aim. It can’t distinguish the top-tier students aiming for medical school,” the teacher said. “Math filters students for medical schools, and Korean determines who gets into medical schools in Seoul. Those decisions can hinge on just one or two questions.”

“If there are too many perfect scorers, that means we failed to differentiate student performance. That failure is just as serious as an error in the question,” another professor said.

“The media often asks, ‘Why give students extremely difficult questions that even university professors can’t solve?’ When I try CSAT questions in my own field, I often have to think for quite a while. The better the question, the more this happens. Yet some students can solve these in under two minutes. As question writers, we must distinguish those very top students from the ones just below them, so we have no choice but to include these super-high‑difficulty questions.”

Oh Seung-geol, president of the Korea Institute for Curriculum and Evaluation (KICE), announces the scoring results of the 2026 College Scholastic Ability Test (CSAT) at the Ministry of Education in Government Complex Sejong in Sejong City on Dec. 4, 2025. Yonhap

Mid‑ and lower‑tier students often view the CSAT as a “league only for the top tier” because of its relentless difficulty.

Song, a 26-year-old pharmacy student outside Seoul, described the reality, saying “When I tutor lower‑ranking students, I feel there are almost no CSAT questions they can solve. Students say they lose the will to continue because they can’t even finish half the exam.

The era of 'sub-killer' questions

When politics intervenes in education without a clear philosophy, the entire field is thrown into confusion. A prime example came in June 2023, when former President Yoon Suk Yeol ordered the CSAT committee — with less than six months before the exam — to “exclude killer questions.”

Critics immediately pointed out the contradiction. Removing killer questions while preserving the exam’s role in differentiating students would inevitably backfire, experts warned, as the overall number of high-difficulty questions would simply increase.

Another issue was the government’s vague definition of what counts as a killer question. As insiders pointed out, a “difficult question” is not the same thing as a “bad question.”

These fears became reality. The void left by the ban was quickly filled by a larger volume of "sub-killer questions."

Sung Ki‑sun, a professor at the Catholic University of Korea and former KICE president from 2017 to 2021, described this as a classic “balloon effect.”

“Even if the super‑high‑difficulty questions disappear, the exam committee still has to maintain the overall difficulty level. So the number of high‑difficulty questions — previously around seven — increased to twelve or thirteen,” he said. “Students who expected an easier CSAT were frustrated because it turned out to be harder than they thought.”

Indeed, the 2024 CSAT, the first test administered after Yoon’s order, became one of the most difficult exams on record, with only one perfect scorer.

Instructors analyze questions, including so-called "killer questions," at Jongro Academy in Yangcheon-gu, Seoul, on Sept. 6, 2023, the day the September mock evaluation for the 2024 College Scholastic Ability Test (CSAT) was administered. Korea Times photo by Wang Tae-seog

Critics argue that eliminating killer questions has produced more “uneducational” questions. Instead of assessing essential knowledge through well‑crafted questions, the exam increasingly relies on trivial details to sort students.

“I can see that wordplay‑based or overly twisted questions have increased noticeably,” the question writer for the Korean‑language section said. “Now I have to teach students to check each character’s name one by one and confirm whether someone ‘said’ something or merely ‘whispered’ it.”

Jeong admitted to this tactical shift. “If we think a question will be condemned as a killer question, we slightly lower its difficulty but twist the logic so it takes longer to solve,” he said.

Seong Tae‑je, who was a former KICE president from 2011 to 2014, criticized the growing use of “attractive distractors.”

“Attractive distractors were originally intended to reveal what students misunderstand or fail to grasp clearly, helping them correct their learning,” he said. “But at some point, they became simple traps designed to cause mistakes.”

English: relative grading in disguise

It has long been argued that the CSAT English section creates confusion by operating as “absolute grading in name only.”

Although the system was originally intended to ease study burdens and curb private education costs, last year’s English exam came under fire for its extreme difficulty. Only 3.11 percent of students earned Grade one. This is the lowest figure since the shift to absolute grading in 2018, and even below the 4-percent threshold under the previous relative‑grading system.

The fluctuations in difficulty are another major problem. In KICE’s June mock test last year, 19.1 percent of students earned Grade one, but the figure plummeted to 4.50 percent in the September test.

Lee Byung‑min, a professor of English education at Seoul National University who first proposed adopting absolute grading, criticized the gap between theory and practice.

“In a true absolute grading system, everyone who meets the standard should receive Grade one, even if that’s half of all students,” he said. “But CSAT English is nothing more than relative grading disguised as absolute grading, where the Grade-one quota is merely slightly expanded."

In reality, KICE targets a Grade one ratio of 7 to 9 percent and adjusts difficulty accordingly, refusing to give up question differentiation.

“The difference is just one point. A student with a 90 gets Grade 1, while a student with 89 gets Grade 2. That’s unfair,” Lee said.

With the Grade‑one ratio hitting record lows last year, conspiracy theories began circulating that an “English‑education lobbyist” pressured the exam committee to raise the difficulty. Since the introduction of absolute grading, some English‑major communities and private‑education insiders have consistently argued for a return to a relative grading system, claiming, "students' focus in English classes has declined, and demand for English tutoring has dropped."

Question 24 of the English section of the 2026 CSAT, which received approximately 370 formal objections. Captured from KICE's webiste

The paradox of EBS questions

This phenomenon is particularly pronounced in the Korean-language and English sections. Because students have already studied passages from EBS, a state-run public educational broadcaster, question writers feel compelled to “twist” the questions significantly during the review process to maintain question differentiation.

“We can’t let students say, ‘I studied EBS thoroughly and it didn’t help,’” a teacher from the exam committee said. “But if we simply use the passage as is, the question loses question differentiation. We have to twist the question so the correct answer doesn’t become obvious just because students are familiar with the text.”

The most difficult questions in last year’s CSAT Korean-language section (questions 10–12) were questions adapted from EBS workbooks. They used an EBS passage on "alloys with low thermal‑expansion coefficients." Question 12 recorded an accuracy rate of just 22.3 percent among EBS users — the actual overall rate was likely even lower.

Kim Chang‑won, a former chair of the exam committee, argues that the current approach has reached its “technical limit.”

“To meet the mandated proportion of EBS‑sourced questions across the three major tests each year, we are scraping the bottom of the barrel for usable passages,” Kim said. “But cram schools also analyze EBS materials using AI to generate hundreds of predicted variations. If we use the same passages, the questions we made inevitably resemble those predictions. To avoid that overlap, questions tied to EBS textbooks end up becoming trickier.”

EBS textbooks aligned with the College Scholastic Ability Test (CSAT) are displayed at the Gwanghwamun branch of Kyobo Book Centre in Seoul. Yonhap

Kim Sung-hoon, a professor emeritus at Dongguk University who led KICE in 2014, explained that the policy inadvertently spawned a new niche in the private sector known as the EBS analysis industry, which focuses on predicting passages likely to appear on the exam.

Critics point out that only the mathematics section has remained true to the policy’s original goal of reducing academic burdens.

According to one exam‑committee professor, math questions sourced from EBS are intentionally kept at low‑to‑medium difficulty to ensure that even weaker students can secure a minimum score by studying EBS materials.

The rising accuracy of “predicted questions” from private education companies has been also a headache for the exam committee. The CSAT team reviews 400 to 500 commercial workbooks — including so‑called “Final” editions published while the committee is in quarantine — and removes any question similar to a commercial question, even at the last minute.

Test makers call this "being hit by a duplicate." They must discard and recreate any overlapping question, even if it has already been fully approved.

“If a question identical to a cram‑school question appears on the CSAT, its ‘hit’ is advertised the next year,” an exam‑committee teacher said. “KICE considers that extremely humiliating.”

KICE admitted that last year’s exceptionally difficult English exam resulted partly from the last-minute replacement of multiple questions deemed too similar to commercial mock tests. Former KICE head Oh explained, “We failed to review the difficulty closely because we were rushing to replace questions that overlapped with commercial materials.”

It is also ironic that while the CSAT relies heavily on EBS workbooks, standard school textbooks are often deliberately sidelined. Question writers intentionally avoid passages found in specific textbooks to prevent fairness disputes.

Expiration date of the 'perfect exam'

Why, then, does the CSAT continue to hold such authority as the decisive exam for university admissions? The Korean‑language question writer on the exam committee pointed to the widespread belief that “among multiple‑choice tests, the CSAT has the highest level of technical perfection in the world.”

Whenever corruption scandals erupt over GPA-based admissions, the CSAT regains standing as the “most fair” and quantifiable standard, where hundreds of thousands of students solve the same questions at the same time.

A student checks her report card for the 2026 College Scholastic Ability Test (CSAT) at Daegu Girls' High School in Daegu on Dec. 5, 2025, the day scores were distributed. Yonhap

However, critics are increasingly questioning whether this multiple-choice format is suitable for measuring future capabilities.

“We are expected to perform the impossible mission — avoiding textbook passages, matching the mandated EBS sourcing rate, maintaining question differentiation, keeping difficulty at the appropriate level without killer questions, and ensuring that none of the questions are overlapping with those found in private‑education materials,” said former committee chair Kim.

His explanation underscores a system pushed to its limits.

After last year’s exam, several media outlets reported that generative AI models such as ChatGPT saw their CSAT math scores jump from Grade 9 to Grade 1 in just two years, completing the test in only fifteen minutes. For many observers, it signaled that the CSAT’s expiration date is fast approaching.

This article from the Hankook Ilbo, the sister publication of The Korea Times, is translated by a generative AI system and edited by The Korea Times.