StudyDaily
Medicine, biology and nutrition
Open questions on clinical practice, anatomy, human physiology and genetics, cell biology, evolution and nutrition, drawn from the MMLU benchmark (Hendrycks et al., 2021), which gathers multiple-choice items of varied origin. Each card asks one question and answers it in full — no multiple-choice letters.
61 cartes EN MIT
Téléchargement direct de l’archive ; la licence et l’attribution sont incluses dans le fichier.
Cinq cartes de ce paquet
Extraites de l’archive publiée, sans retouche. Une fiche de paquet qui ne montre aucune carte ne dit rien.
-
Recto Which muscles are normally active during quiet inspiration?Verso External intercostal muscles and diaphragm
-
Recto In what situation are closed pouches applied?Verso The patient has a colostomy
-
Recto How do cyanobacteria differ from green unicellular algae?Verso Cyanobacteria have no nuclei
-
Recto During muscular contraction, interactions between myosin and actin allow for shortening of each sarcomere. In addition to the power stroke, what other process of muscle contraction requires ATP?Verso Myosin-actin detachment
-
Recto A man contracts the same flu strain for the second time in a single winter season. The second time he experiences fewer symptoms and recovers more quickly. Which cells are responsible for this rapid recovery?Verso Memory cells
Licence, source et attribution
Découvrez l’origine de ce paquet et les conditions pour l’utiliser ou le partager.
- Licence
-
MIT License
MIT - Attribution
- Source : https://people.eecs.berkeley.edu/~hendrycks/data.tar — Licence : MIT License (https://opensource.org/licenses/MIT)
- Source
- MMLU — Measuring Massive Multitask Language Understanding (Hendrycks et al., 2021)
- Récupéré le
- Thème
- Biologie et médecine
- Langue
- English
- Version
- 3
- Mis à jour le
Ce paquet a été modifié lors de son intégration : Every MMLU question is a multiple-choice item: four options and a letter for an answer. Memoquiz resolved the letter, wrote the correct option out in full on the back and threw the other three away — a card answering “B” would teach nothing. Questions that do not stand up without their options were dropped: anything pointing at material the card does not carry (a figure, a table, a passage, a chapter, a scenario), anything naming the answer grid itself (“which of the following”, “which one of these”, bare “Which is…”), every comparative superlative (“best”, “most likely”, “least appropriate”), negative questions, questions whose answer is a number, a calculation, a true/false or a rank, and time-stamped wording (“currently”, “as of 2020”, “in the last two decades”). Whole subjects were left out for the same reason — calculation subjects, multi-paragraph case studies, dated world statistics, virology and the US general-trivia subject, which carries factual errors and extraction damage. What survived the filters was then read one question at a time, and what read badly was named and removed in the connector source. Retained questions are reproduced word for word; only mechanical repairs are applied — whitespace normalised, raw ligatures decomposed, flattened chemical formulas and units given their subscripts back (H2O → H₂O, kg/m2 → kg/m²), the trailing full stop of answers removed so punctuation does not change from card to card, and a capital restored on answers the source left in lower case. No word is rewritten.
Droits des médias, fichier par fichier
Ce paquet ne contient ni image ni son : il est uniquement textuel.