A Comparative Study of Open-Source Large Language Models, GPT-4 and Claude 2: Multiple-Choice Test Taking in Nephrology
Paper • 2308.04709 • Published
Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code: DatasetGenerationCastError
Exception: DatasetGenerationCastError
Message: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 7 new columns ({'Subject', 'Context', 'Question', 'Solution', 'ID', 'Choices', 'Answer'}) and 2 missing columns ({'explanation', 'id'}).
This happened while the csv dataset builder was generating data using
hf://datasets/SeanWu25/NEJM-AI_Benchmarking_Medical_Language_Models/NEJM_All_Questions_And_Answers.csv (at revision 7839934224e0c50e50143d98f8db1d56462b87b1)
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback: Traceback (most recent call last):
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 2011, in _prepare_split_single
writer.write_table(table)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/arrow_writer.py", line 585, in write_table
pa_table = table_cast(pa_table, self._schema)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2302, in table_cast
return cast_table_to_schema(table, schema)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/table.py", line 2256, in cast_table_to_schema
raise CastError(
datasets.table.CastError: Couldn't cast
ID: int64
Context: string
Question: string
Choices: string
Answer: string
Solution: string
Subject: string
-- schema metadata --
pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1048
to
{'id': Value(dtype='int64', id=None), 'explanation': Value(dtype='string', id=None)}
because column names don't match
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1321, in compute_config_parquet_and_info_response
parquet_operations = convert_to_parquet(builder)
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 935, in convert_to_parquet
builder.download_and_prepare(
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 1027, in download_and_prepare
self._download_and_prepare(
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 1122, in _download_and_prepare
self._prepare_split(split_generator, **prepare_split_kwargs)
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 1882, in _prepare_split
for job_id, done, content in self._prepare_split_single(
File "/src/services/worker/.venv/lib/python3.9/site-packages/datasets/builder.py", line 2013, in _prepare_split_single
raise DatasetGenerationCastError.from_cast_error(
datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 7 new columns ({'Subject', 'Context', 'Question', 'Solution', 'ID', 'Choices', 'Answer'}) and 2 missing columns ({'explanation', 'id'}).
This happened while the csv dataset builder was generating data using
hf://datasets/SeanWu25/NEJM-AI_Benchmarking_Medical_Language_Models/NEJM_All_Questions_And_Answers.csv (at revision 7839934224e0c50e50143d98f8db1d56462b87b1)
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
id int64 | explanation string |
|---|---|
2,464 | Answer D: About 50% of people with hypertension are controlled |
2,465 | Answer A: Increased long-term BP variability is associated with an increased risk of cardiovascular events despite adequate BP control on most visits |
2,466 | Answer D: Low dietary potassium promotes hypertension by increasing the activity of the sodium-chloride cotransporter (on a high sodium diet) |
2,467 | Answer D: Add amiloride |
2,468 | Answer B: In 1 year |
2,469 | Answer B: Obtain 24-hour ambulatory BP monitoring |
2,470 | Answer D: Add spironolactone |
2,471 | Answer B: Home BP tele-monitoring supervised by a health care provider results in higher rates of BP control than usual care Home BP monitoring, with or without remote telemonitoring, is associated with better BP control; therefore, choice B is correct, and choice A is incorrect. The best results occur when a healthca... |
2,472 | Answer D: 110/75 mmHg |
2,473 | Answer A: Droxidopa |
2,474 | Answer D: Make no changes to the current regimen |
2,475 | Answer D: Her BMI |
2,476 | Answer C: Inappropriately low atrial natriuretic peptide levels |
2,477 | Answer D: Regular use of continuous positive airway pressure for at least 4 hours per night Patients with obesity have an increased incidence of both sleep apnea and resistant hypertension. A recent study has shown that the combination of weight loss with regular use of CPAP improves BP and can reverse abnormal dippin... |
2,478 | Answer D: Order a CT angiogram |
2,479 | Answer D: Refer for adrenal vein sampling |
2,480 | Answer C: A lateralization index of 6 |
2,481 | Answer D: Plasma metanephrines |
2,482 | Answer C: Genetic testing |
2,483 | Answer D: Maintenance of his systolic BP at <120 mmHg |
2,484 | Answer C: Reversal of INR to <1.3 plus lowering of systolic BP <160 mmHg within 4 hours |
2,485 | Answer D: The absence of nocturnal dipping |
2,486 | Answer B: Underlying CKD |
2,488 | Answer D: Add amlodipine, targeting a systolic BP of <120 mmHg |
2,489 | Answer C: The low BP target (120 mmHg) resulted in a 38% reduction in the risk of heart failure In SPRINT, a systolic BP target of <120 mmHg, compared to a target of 140 mmHg, resulted in lower risk of several cardiovascular endpoints as well as all-cause death. Specifically, the composite endpoint of myocardial infar... |
2,490 | Answer C: Refer for percutaneous renal angioplasty |
2,491 | Answer E: 25OHD replacement has no effect on BP in hypertensive patients with 25OHD deficiency |
2,492 | Answer C: Beetroot juice lowers ambulatory BP by about 8/5 mmHg compared with placebo Beetroot juice is a good source of nitrates, resulting in BP lowering through the provision of nitrates that result in vasodilatation, improved endothelial function, and lower BP (-7.7/5.2 mmHg on 24h ABPM). There is no known effect ... |
2,493 | Answer B: Target the hemoglobin A1c to <7.0 % before conception |
2,494 | Answer C: Increased AST and ALT |
2,495 | Answer E: Make no changes to the current regimen |
2,496 | Answer A: Induction of labor and delivery |
2,497 | Answer D: Continue intravenous magnesium and transition to oral antihypertensive medications |
2,498 | Answer A: Labetalol |
2,499 | Answer C: Aspirin at 81 mg daily |
2,500 | Answer C: It increases the risk of cardiovascular disease |
2,501 | Answer B: She is at increased risk of urinary tract infection |
2,502 | Answer D: Acute fatty liver of pregnancy (AFLP) |
2,503 | Answer A: Transfuse platelets and packed red blood cells |
2,504 | Answer B: Long-chain 3-hydroxyl CoA dehydrogenase |
2,505 | Answer C: Observation with expectant medical management |
2,506 | Answer D: Renal blood flow returns to pre-pregnancy levels |
2,507 | Answer D: Preeclampsia |
2,508 | Answer A: Intravenous normal saline plus furosemide |
2,509 | Answer D: Hemodialysis totaling ≥36 h/wk |
2,510 | Answer C: A low soluble fms-like tyrosine kinase-1 to placental growth factor ratio |
2,511 | Answer C: Enalapril |
2,512 | Answer B: She should start aspirin 81 mg daily immediately after pregnancy is diagnosed Approximately one in three women who have received a kidney transplant develop preeclampsia during pregnancy. Therefore, prophylaxis with low-dose aspirin is appropriate. In addition, heightened antenatal surveillance should be off... |
2,513 | Answer A: The probability of achieving pregnancy is reduced compared with spontaneous conception |
2,514 | Answer B: Stop mycophenolate mofetil and start azathioprine |
2,515 | Answer D: Anti-Ro and anti-La antibodies |
2,516 | Answer C: Furosemide |
2,517 | Answer B: Reduce the tacrolimus dose |
2,518 | Answer B: Enzyme replacement therapy |
2,519 | Answer D: It does not increase the risk for a small gestation baby |
2,520 | Answer C: A reduction in the loading dose and maintenance dose |
2,521 | Answer C: Empiric therapy with predisone and azathioprine, followed by adjusted treatment after delivery based on kidney biopsy performed postpartum |
2,522 | Answer A: Her risk of progression to ESRD during or shortly postpartum is about 20% |
2,523 | Answer D: IgA nephropathy |
2,524 | Answer C: Measure or estimate 24-hour creatinine excretion rate |
2,525 | Answer E: Rituximab |
2,526 | Answer D: Stain prior kidney biopsy with Congo Red |
2,527 | Answer E: No additional diagnostic studies are required |
2,528 | Answer A: Oral angiotensin receptor blockers (ARBs)d |
2,529 | Answer A: Oral glucocorticoids |
2,530 | Answer B: Oral cyclosporine and low-dose steroids for ≥4 months |
2,531 | Answer C: As in A plus widespread effacement of the podocyte foot processes by electron microscopy |
2,532 | Answer C: Hyperexpression of PLA2R antigen by immunohistochemical studies of pronasedigested, paraffin-preserved specimens |
2,533 | Answer E: Continued supportive care and observation |
2,534 | Answer B: Pronase digestion of paraffin-embedded material and staining for monoclonal IgG (light and heavy chains) |
2,535 | Answer A: Sarcoidosis causing secondary membranous nephropathy |
2,536 | Answer B: Oral ACEIs or ARBs |
2,537 | Answer D: Perform a laser dissection/mass spectrometry analysis of the biopsy specimen |
2,538 | Answer E: Serum immunofixation and free light-chain assays |
2,539 | Answer B: An underlying autoimmune disorder |
2,540 | Answer B: Staining the kidney biopsy for hepatitis B surface antigen and core antigen |
2,541 | Answer D: Sofosbuvir plus simeprevir plus rituximab |
2,542 | Answer C: IgA−dominant, infection−related GN |
2,543 | Answer C: Kidney prognosis is likely to be good long term, but uncertainty exists; therefore, yearly follow-up is advisable |
2,544 | Answer A: A calcineurin inhibitor |
2,545 | Answer C: Treat with cyclophosphamide and corticosteroids as outlined in choice B but reduce the doses by ≥20% |
2,546 | Answer B: Either intravenous cyclophosphamide or intravenous rituximab |
2,547 | Answer A: Rituximab > azathioprine > mycophenolate > extended cyclophosphamide |
2,548 | Answer A: Plasmapheresis, high−dose corticosteroids, and cyclophosphamide |
2,549 | Answer A: Provide supportive care and management for ESRD |
2,550 | Answer A: The long−term prognosis is favorable, because the proteinuria declined to <0.7 g/d after 1 year |
2,551 | Answer B: Add low−dose Euro lupus cyclophosphamide for 3 months to corticosteroids |
2,552 | Answer B: A sustained remission for at least 24 months |
2,553 | Answer E: Conduct a detailed evaluation of his diet |
2,554 | Answer E: Adynamic bone disease |
2,555 | Answer B: All surgical approaches, such as partial, subtotal, or total parathyroidectomy with or without autotransplantation, are effective |
2,556 | Answer B: Begin cinacalcet therapy |
2,557 | Answer D: Cinacalcet |
2,558 | Answer D: Add teriparatide |
2,559 | Answer A: The effect of phosphate binders does not increase linearly with dose |
2,560 | Answer C: Lithium-related primary hyperparathyroidism |
2,561 | Answer C: Elevated IgGκ monoclonal protein |
2,564 | Answer D: Add sodium thiosulfate |
2,566 | Answer A: Start bisphosphonate therapy |
2,567 | Answer C: Identify the method of serum phosphate measurement |
End of preview.
A Comparative Study of Open-Source Large Language Models
Dataset Overview
Welcome to the dataset repository for our paper, "A Comparative Study of Open-Source Large Language Models, GPT-4 and Claude 2: Multiple-Choice Test Taking in Nephrology." The preprint of the paper can be accessed here.
Files
This repository contains two key files:
NEJM_All_Questions_And_Answers.csv: This file includes all the questions and corresponding answers used in the study.
Ground_Truth_Answers.csv: This file provides ground truth explanations associated with the questions in the main dataset.
Usage
To utilize this dataset for your research or experimentation:
- Download: Obtain the dataset files from this repository.
- Load: Import the dataset into your preferred data analysis or machine learning environment.
- Explore: Investigate the questions, answers, and ground truth explanations for your specific use case.
Paper
Our paper is accepted to NEJM-AI. For now please read the pre-print at the link: https://arxiv.org/abs/2308.04709
- Downloads last month
- 17