PatientSafeBench: Evaluating the Safety of Medical LLMs for Patient Use
Bibliographic record
Abstract
Large Language Models (LLMs) in the medical domain have been primarily developed and validated for healthcare professionals, leaving a significant gap in patient-centered adaptation. As real-world patient use of these models poses safety risks, rigorous evaluation tailored for patient interaction scenarios becomes essential. To address this, we introduce PatientSafeBench, a novel benchmark assessing both the safety and utility of LLMs in patient-facing contexts. It comprises five categories and 25 subcategories, each representing critical aspects of LLM performance for patient use. We developed 500 evaluation queries grounded in real clinical cases, with scoring criteria reviewed by four medical professionals. We evaluated 11 different LLMs on PatientSafeBench using a multi-judge approach, scoring responses on a 10-point scale with hierarchical safety thresholds. The results reveal that no model met our safety criteria for patient use, with medical-specific LLMs surprisingly underperforming general-purpose models. All models showed consistent weaknesses in temporal relevance, transparency, personalization, and user engagement. These findings highlight the need for dedicated patient-centered benchmarks to ensure the safety and effectiveness of LLMs in patient-facing applications.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.021 | 0.128 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
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 source (direct Gemma or distilled Codex), 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".