Comparing the Performance of ChatGPT and GPT-4 versus a Cohort of Medical Students on an Official University of Toronto Undergraduate Medical Education Progress Test
Bibliographic record
Abstract
A bstract Background Large language model (LLM) based chatbots have recently received broad social uptake; demonstrating remarkable abilities in natural language understanding, natural language generation, dialogue, and logic/reasoning. Objective To compare the performance of two LLM-based chatbots, versus a cohort of medical students, on a University of Toronto undergraduate medical progress test. Methods We report the mean number of correct responses, stratified by year of training/education, for each cohort of undergraduate medical students. We report counts/percentages of correctly answered test questions for each of ChatGPT and GPT-4. We compare the performance of ChatGPT versus GPT-4 using McNemar’s test for dependent proportions. We compare whether the percentage of correctly answered test questions for ChatGPT or GPT-4 fall within/outside the confidence intervals for the mean number of correct responses for each of the cohorts of undergraduate medical education students. Results A total of N=1057 University of Toronto undergraduate medical students completed the progress test during the Fall-2022 and Winter-2023 semesters. Student performance improved with increased training/education levels: UME-Year1 mean=36.3%; UME-Year2 mean=44.1%; UME-Year3 mean=52.2%; UME-Year4 mean=58.5%. ChatGPT answered 68/100 (68.0%) questions correctly; whereas, GPT-4 answered 79/100 (79.0%) questions correctly. GPT-4 performance was statistically significantly greater than ChatGPT (P=0.034). GPT-4 performed at a level equivalent to the top performing undergraduate medical student (79/100 questions correctly answered). Conclusions This study adds to a growing body of literature demonstrating the remarkable performance of LLM-based chatbots on medical tests. GPT-4 performed at a level comparable to the best performing undergraduate medical student who attempted the progress test in 2022/2023. Future work will investigate the potential application of LLM-chatbots as tools for assisting learners/educators in medical education.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".