Dataset Viewer
Auto-converted to Parquet Duplicate
prompt
stringlengths
199k
227k
category
stringclasses
1 value
extra_info
dict
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"D","answer_text":"Vincristine"},"scoring_function":"longhealth_acc(...TRUNCATED)
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"D","answer_text":"She has 3-vessel CHD."},"scoring_function":"long(...TRUNCATED)
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"A","answer_text":"Multiple sclerosis"},"scoring_function":"longhea(...TRUNCATED)
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"B","answer_text":"0.5 mg, 1 capsule at bedtime"},"scoring_function(...TRUNCATED)
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"D","answer_text":"6 cycles of R-Pola-CHP"},"scoring_function":"lon(...TRUNCATED)
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"C","answer_text":"Thoracic vertebra 10"},"scoring_function":"longh(...TRUNCATED)
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"A","answer_text":"Omnipaque 320"},"scoring_function":"longhealth_a(...TRUNCATED)
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"B","answer_text":"MR Brain > MR thoracic/lumbar spine > CT chest/a(...TRUNCATED)
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"E","answer_text":"The patient started taking 2mg of Melatonin in t(...TRUNCATED)
"<|im_start|>system\nYou are a medical expert assistant analyzing clinical documents. Read the patie(...TRUNCATED)
memwrap/longhealth
{"ground_truth":{"answer_letter":"B","answer_text":"The patient's pain management was escalated, as (...TRUNCATED)
End of preview. Expand in Data Studio

LongHealth5-CoT

Same 5-patient concatenated LongHealth benchmark as leonli66/longhealth5, but the final instruction replaces Respond with ONLY the letter... with:

Please think step by step very briefly and then, on a new line, respond with the letter (A, B, C, D, or E) of the correct option. If you are not sure, still make a guess.

From the Zweiger et al. compaction paper — gives hybrid-thinking and memwrap models a chance to reason before committing to a letter.

400 questions, 4 groups of 5 patients, ~60K tokens per context. Scored with longhealth_accuracy.

Downloads last month
37
Free AI Image Generator No sign-up. Instant results. Open Now