The fact that the valedictorian of the thinking assessment (TSA) exam phase 2 in 2026 of Hanoi University of Science and Technology achieved 98.98/100 points despite being wrong in 9 questions is attracting attention from candidates and parents.
Before these concerns, Assoc. Prof. Dr. Vu Duy Hai - Head of Admissions - Career Guidance Department, Hanoi University of Science and Technology - shared with the press to clarify the scoring mechanism according to the modern education measurement model, instead of the traditional calculation method.
Total scores are not only based on the total number of correct answers
According to Assoc. Prof. Dr. Vu Duy Hai, for many years, exams in Vietnam have used a rough scoring method, meaning each correct question is calculated with the same score and added to get the final result.
This approach is simple but reveals limitations when it does not distinguish the level of competence between candidates with the same number of correct questions, especially in cases where the questions are different in difficulty.
In fact, two candidates who achieved the same score may have correctly answered different groups of questions, with equivalent difficulty levels. At that time, scores do not accurately reflect competence, and at the same time cause difficulties in comparing results between exam questions or exam phases with different levels.
To overcome this, Hanoi University of Science and Technology applies the IRT (Item Response Theory) model, a modern educational measurement method used in many international exams.
The TSA exam uses a 2-parameter IRT model, including the difficulty and differentiation of the questions. Accordingly, each question in the exam is assigned its own parameters, reflecting the difficulty level and ability to classify candidates. At the same time, each candidate is also estimated a certain level of competence through the entire test.
Therefore, candidates who do many difficult questions correctly will have an advantage in points. While candidates who only do easy questions correctly will have a lower score.
4-step IRT score process with 2 parameters in TSA
The scoring process according to the 2-parameter IRT model is carried out in four steps.
Step 1: Based on the candidate's test data, the grading software will estimate the difficulty and differentiation of each question in the current exam.
Step 2: The grading software estimates the thinking ability score of each candidate based on the test results and the scale of the current exam.
Step 3: Convert the scale of the current exam to a common reference scale with previous exams to bring the candidate's thinking ability score to the same scale.
Step 4: Convert the thinking ability score of the referenced candidate to the TSA scale.
Assoc. Prof. Dr. Vu Duy Hai said that grading according to IRT helps to assess more accurately and fairly, when the score is not only based on the number of correct questions but also the result of the process of assessing competence based on the entire work.
At the same time, the scale of the exam questions will be converted to a common reference scale between exam periods to ensure equivalence and fairness.
The application of this measurement model is considered a step forward in enrollment innovation, approaching international exams, but candidates need to understand clearly to avoid confusion when comparing scores.
