Comparison of <scp>ChatGPT</scp> ‐4, Copilot, Bard and Gemini Ultra on an Otolaryngology Question Bank
Bibliographic record
Abstract
OBJECTIVE: To compare the performance of Google Bard, Microsoft Copilot, GPT-4 with vision (GPT-4) and Gemini Ultra on the OTO Chautauqua, a student-created, faculty-reviewed otolaryngology question bank. STUDY DESIGN: Comparative performance evaluation of different LLMs. SETTING: N/A. PARTICIPANTS: N/A. METHODS: Large language models (LLMs) are being extensively tested in medical education. However, their accuracy and effectiveness remain understudied, particularly in otolaryngology. This study involved inputting 350 single-best-answer multiple choice questions, including 18 image-based questions, into four LLMS. Questions were sourced from six independent question banks related to (a) rhinology, (b) head and neck oncology, (c) endocrinology, (d) general otolaryngology, (e) paediatrics, (f) otology, (g) facial plastics, reconstruction and (h) trauma. LLMs were instructed to provide an output reasoning for their answers, the length of which was recorded. RESULTS: Aggregate and subgroup analysis revealed that Gemini (79.8%) outperformed the other LLMs, followed by GPT-4 (71.1%), Copilot (68.0%), and Bard (65.1%) in accuracy. The LLMs had significantly different average response lengths, with Bard (x̄ = 1685.24) being the longest and no difference between GPT-4 (x̄ = 827.34) and Copilot (x̄ = 904.12). Gemini's longer responses (x̄ =1291.68) included explanatory images and links. Gemini and GPT-4 correctly answered image-based questions (n = 18), unlike Copilot and Bard, highlighting their adaptability and multimodal capabilities. CONCLUSION: Gemini outperformed the other LLMs in terms of accuracy, followed by GPT-4, Copilot and Bard. GPT-4, although it has the second-highest accuracy, provides concise and relevant explanations. Despite the promising performance of LLMs, medical learners should cautiously assess accuracy and decision-making reliability.
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 | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| 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".