Merge pull request #922 from EleutherAI/mmlu_subgroups

[Refactor] Mmlu subgroups and weight avg

lm_eval/tasks/mmlu/flan_cot_fewshot/mmlu_virology.yaml

-dataset_name: virology
-description: 'The following are multiple choice questions (with answers) about virology.
-
-
-  Q: The median survival time to AIDS and death was established by following:
-
-  (A) Seroprevalent HIV-infected individuals (B) Seronegatives (C) Seroconverters
-  (D) High-risk seronegatives
-
-  A: Let''s think step by step. We refer to Wikipedia articles on virology for help.
-  The median survival time to AIDS and death was established as a result of the development
-  of seroconverters. The answer is (C).
-
-
-  Q: Which of the following is a morphological characteristic of the paramyxoviruses.
-
-  (A) Fragile viruses often visualised with RNA spewing from the inside (B) Elongate
-  viruses (C) Icosahedral viruses with envelope (D) Very large viruses
-
-  A: Let''s think step by step. We refer to Wikipedia articles on virology for help.
-  Paramyxoviruses are fragile viruses often visualised with RNA spewing from the inside.
-  The answer is (A).
-
-
-  Q: The most important goal of a behavioral intervention is:
-
-  (A) Change in behavior (B) Comprehensive coverage (C) Effective use of behavioral
-  theory (D) Sustained behavior change
-
-  A: Let''s think step by step. We refer to Wikipedia articles on virology for help.
-  The prim goal of a behavioral intervention is to cause sustained behavior change.
-  The answer is (D).
-
-
-  Q: A key factor facilitating the application of nested case-control studies from
-  the MACS was:
-
-  (A) Data collection (B) Establishment of a repository of biologic specimens (C)
-  Participant interest (D) Administration of the questionnaire by staff
-
-  A: Let''s think step by step. We refer to Wikipedia articles on virology for help.
-  The Multicenter AIDS Cohort Study''s use of nested case-control studies was facilitated
-  by the establishment of a repository of biologic specimens. The answer is (B).
-
-
-  Q: Why are parvoviruses a highly impactful parasite?
-
-  (A) Because they have no nucleic acid (B) They require a helper virus (C) Only replicate
-  in dividing cells (D) Can integrate into host chromosomes
-
-  A: Let''s think step by step. We refer to Wikipedia articles on virology for help.
-  Paroviruses are highly impactful because they do not have nucleic acid. The answer
-  is (A).'
-include: _mmlu_flan_cot_fewshot_template_yaml
-task: mmlu_flan_cot_fewshot_virology
+"dataset_name": "virology"
+"description": "The following are multiple choice questions (with answers) about virology.\n\
+  \nQ: The median survival time to AIDS and death was established by following:\n\
+  (A) Seroprevalent HIV-infected individuals (B) Seronegatives (C) Seroconverters\
+  \ (D) High-risk seronegatives\nA: Let's think step by step. We refer to Wikipedia\
+  \ articles on virology for help. The median survival time to AIDS and death was\
+  \ established as a result of the development of seroconverters. The answer is (C).\n\
+  \nQ: Which of the following is a morphological characteristic of the paramyxoviruses.\n\
+  (A) Fragile viruses often visualised with RNA spewing from the inside (B) Elongate\
+  \ viruses (C) Icosahedral viruses with envelope (D) Very large viruses\nA: Let's\
+  \ think step by step. We refer to Wikipedia articles on virology for help. Paramyxoviruses\
+  \ are fragile viruses often visualised with RNA spewing from the inside. The answer\
+  \ is (A).\n\nQ: The most important goal of a behavioral intervention is:\n(A) Change\
+  \ in behavior (B) Comprehensive coverage (C) Effective use of behavioral theory\
+  \ (D) Sustained behavior change\nA: Let's think step by step. We refer to Wikipedia\
+  \ articles on virology for help. The prim goal of a behavioral intervention is to\
+  \ cause sustained behavior change. The answer is (D).\n\nQ: A key factor facilitating\
+  \ the application of nested case-control studies from the MACS was:\n(A) Data collection\
+  \ (B) Establishment of a repository of biologic specimens (C) Participant interest\
+  \ (D) Administration of the questionnaire by staff\nA: Let's think step by step.\
+  \ We refer to Wikipedia articles on virology for help. The Multicenter AIDS Cohort\
+  \ Study's use of nested case-control studies was facilitated by the establishment\
+  \ of a repository of biologic specimens. The answer is (B).\n\nQ: Why are parvoviruses\
+  \ a highly impactful parasite?\n(A) Because they have no nucleic acid (B) They require\
+  \ a helper virus (C) Only replicate in dividing cells (D) Can integrate into host\
+  \ chromosomes\nA: Let's think step by step. We refer to Wikipedia articles on virology\
+  \ for help. Paroviruses are highly impactful because they do not have nucleic acid.\
+  \ The answer is (A)."
+"group": "mmlu_flan_cot_fewshot_other"
+"include": "_mmlu_flan_cot_fewshot_template_yaml"
+"task": "mmlu_flan_cot_fewshot_virology"

lm_eval/tasks/mmlu/flan_cot_fewshot/mmlu_world_religions.yaml

-dataset_name: world_religions
-description: 'The following are multiple choice questions (with answers) about world
-  religions.
-
-
-  Q: How can the Upanishads be characterized?
-
-  (A) Ritual texts (B) Philosophical texts (C) Hymns (D) Origin stories
-
-  A: Let''s think step by step. We refer to Wikipedia articles on world religions
-  for help. The Upanishads are the most recent part of Vedas (the oldest scriptures
-  in Hinduism) and supplied the basis of later Hindu philosophy. So they are philosophical
-  texts. The answer is (B).
-
-
-  Q: What is the Second Gem in Buddhism?
-
-  (A) The Dharma (B) The Sangha (C) The Buddha (D) The Bodhisattva
-
-  A: Let''s think step by step. We refer to Wikipedia articles on world religions
-  for help. The Second Gem in Buddhism is The Dharma. The answer is (A).
-
-
-  Q: Which Japanese government promoted a kind of national cult based on the emperor
-  and his associations with kami?
-
-  (A) Honen (B) Tanaka (C) Tokugawa (D) Meiji
-
-  A: Let''s think step by step. We refer to Wikipedia articles on world religions
-  for help. The promotion of a national cult based on the emperor and his associations
-  with Kami happened during the reign of Emperor Meiji (1852-1912). The answer is
-  (D).
-
-
-  Q: In which dynasty was the "Mandate of Heaven" developed to legitimatize the new
-  rulers?
-
-  (A) Shang (B) Zhou (C) Han (D) Xia
-
-  A: Let''s think step by step. We refer to Wikipedia articles on world religions
-  for help. The "Mandate of Heaven" was developed as an ancient Chinese philosophical
-  concept during the Zhou Dynasty (1046-256 BCE). The answer is (B).
-
-
-  Q: What is the sign of the covenant for Jewish males?
-
-  (A) The rainbow (B) Circumcision (C) A son (D) Bar mitzvah
-
-  A: Let''s think step by step. We refer to Wikipedia articles on world religions
-  for help. In Judaism, the most distinctive sign of the covenant is circumcision
-  (brit milah). The answer is (B).'
-include: _mmlu_flan_cot_fewshot_template_yaml
-task: mmlu_flan_cot_fewshot_world_religions
+"dataset_name": "world_religions"
+"description": "The following are multiple choice questions (with answers) about world\
+  \ religions.\n\nQ: How can the Upanishads be characterized?\n(A) Ritual texts (B)\
+  \ Philosophical texts (C) Hymns (D) Origin stories\nA: Let's think step by step.\
+  \ We refer to Wikipedia articles on world religions for help. The Upanishads are\
+  \ the most recent part of Vedas (the oldest scriptures in Hinduism) and supplied\
+  \ the basis of later Hindu philosophy. So they are philosophical texts. The answer\
+  \ is (B).\n\nQ: What is the Second Gem in Buddhism?\n(A) The Dharma (B) The Sangha\
+  \ (C) The Buddha (D) The Bodhisattva\nA: Let's think step by step. We refer to Wikipedia\
+  \ articles on world religions for help. The Second Gem in Buddhism is The Dharma.\
+  \ The answer is (A).\n\nQ: Which Japanese government promoted a kind of national\
+  \ cult based on the emperor and his associations with kami?\n(A) Honen (B) Tanaka\
+  \ (C) Tokugawa (D) Meiji\nA: Let's think step by step. We refer to Wikipedia articles\
+  \ on world religions for help. The promotion of a national cult based on the emperor\
+  \ and his associations with Kami happened during the reign of Emperor Meiji (1852-1912).\
+  \ The answer is (D).\n\nQ: In which dynasty was the \"Mandate of Heaven\" developed\
+  \ to legitimatize the new rulers?\n(A) Shang (B) Zhou (C) Han (D) Xia\nA: Let's\
+  \ think step by step. We refer to Wikipedia articles on world religions for help.\
+  \ The \"Mandate of Heaven\" was developed as an ancient Chinese philosophical concept\
+  \ during the Zhou Dynasty (1046-256 BCE). The answer is (B).\n\nQ: What is the sign\
+  \ of the covenant for Jewish males?\n(A) The rainbow (B) Circumcision (C) A son\
+  \ (D) Bar mitzvah\nA: Let's think step by step. We refer to Wikipedia articles on\
+  \ world religions for help. In Judaism, the most distinctive sign of the covenant\
+  \ is circumcision (brit milah). The answer is (B)."
+"group": "mmlu_flan_cot_fewshot_humanities"
+"include": "_mmlu_flan_cot_fewshot_template_yaml"
+"task": "mmlu_flan_cot_fewshot_world_religions"

lm_eval/tasks/mmlu/flan_cot_zeroshot/_mmlu.yaml

+group: mmlu_flan_cot_zeroshot
+task:
+  - mmlu_flan_cot_zeroshot_stem
+  - mmlu_flan_cot_zeroshot_other
+  - mmlu_flan_cot_zeroshot_social_sciences
+  - mmlu_flan_cot_zeroshot_humanities

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_abstract_algebra.yaml

-dataset_name: abstract_algebra
-description: 'The following are multiple choice questions (with answers) about abstract
-  algebra.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_abstract_algebra
+"dataset_name": "abstract_algebra"
+"description": "The following are multiple choice questions (with answers) about abstract\
+  \ algebra.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_abstract_algebra"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_anatomy.yaml

-dataset_name: anatomy
-description: 'The following are multiple choice questions (with answers) about anatomy.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_anatomy
+"dataset_name": "anatomy"
+"description": "The following are multiple choice questions (with answers) about anatomy.\n\
+  \n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_anatomy"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_astronomy.yaml

-dataset_name: astronomy
-description: 'The following are multiple choice questions (with answers) about astronomy.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_astronomy
+"dataset_name": "astronomy"
+"description": "The following are multiple choice questions (with answers) about astronomy.\n\
+  \n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_astronomy"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_business_ethics.yaml

-dataset_name: business_ethics
-description: 'The following are multiple choice questions (with answers) about business
-  ethics.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_business_ethics
+"dataset_name": "business_ethics"
+"description": "The following are multiple choice questions (with answers) about business\
+  \ ethics.\n\n"
+"group": "mmlu_flan_cot_zeroshot_other"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_business_ethics"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_clinical_knowledge.yaml

-dataset_name: clinical_knowledge
-description: 'The following are multiple choice questions (with answers) about clinical
-  knowledge.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_clinical_knowledge
+"dataset_name": "clinical_knowledge"
+"description": "The following are multiple choice questions (with answers) about clinical\
+  \ knowledge.\n\n"
+"group": "mmlu_flan_cot_zeroshot_other"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_clinical_knowledge"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_college_biology.yaml

-dataset_name: college_biology
-description: 'The following are multiple choice questions (with answers) about college
-  biology.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_college_biology
+"dataset_name": "college_biology"
+"description": "The following are multiple choice questions (with answers) about college\
+  \ biology.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_college_biology"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_college_chemistry.yaml

-dataset_name: college_chemistry
-description: 'The following are multiple choice questions (with answers) about college
-  chemistry.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_college_chemistry
+"dataset_name": "college_chemistry"
+"description": "The following are multiple choice questions (with answers) about college\
+  \ chemistry.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_college_chemistry"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_college_computer_science.yaml

-dataset_name: college_computer_science
-description: 'The following are multiple choice questions (with answers) about college
-  computer science.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_college_computer_science
+"dataset_name": "college_computer_science"
+"description": "The following are multiple choice questions (with answers) about college\
+  \ computer science.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_college_computer_science"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_college_mathematics.yaml

-dataset_name: college_mathematics
-description: 'The following are multiple choice questions (with answers) about college
-  mathematics.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_college_mathematics
+"dataset_name": "college_mathematics"
+"description": "The following are multiple choice questions (with answers) about college\
+  \ mathematics.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_college_mathematics"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_college_medicine.yaml

-dataset_name: college_medicine
-description: 'The following are multiple choice questions (with answers) about college
-  medicine.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_college_medicine
+"dataset_name": "college_medicine"
+"description": "The following are multiple choice questions (with answers) about college\
+  \ medicine.\n\n"
+"group": "mmlu_flan_cot_zeroshot_other"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_college_medicine"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_college_physics.yaml

-dataset_name: college_physics
-description: 'The following are multiple choice questions (with answers) about college
-  physics.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_college_physics
+"dataset_name": "college_physics"
+"description": "The following are multiple choice questions (with answers) about college\
+  \ physics.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_college_physics"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_computer_security.yaml

-dataset_name: computer_security
-description: 'The following are multiple choice questions (with answers) about computer
-  security.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_computer_security
+"dataset_name": "computer_security"
+"description": "The following are multiple choice questions (with answers) about computer\
+  \ security.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_computer_security"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_conceptual_physics.yaml

-dataset_name: conceptual_physics
-description: 'The following are multiple choice questions (with answers) about conceptual
-  physics.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_conceptual_physics
+"dataset_name": "conceptual_physics"
+"description": "The following are multiple choice questions (with answers) about conceptual\
+  \ physics.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_conceptual_physics"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_econometrics.yaml

-dataset_name: econometrics
-description: 'The following are multiple choice questions (with answers) about econometrics.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_econometrics
+"dataset_name": "econometrics"
+"description": "The following are multiple choice questions (with answers) about econometrics.\n\
+  \n"
+"group": "mmlu_flan_cot_zeroshot_social_sciences"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_econometrics"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_electrical_engineering.yaml

-dataset_name: electrical_engineering
-description: 'The following are multiple choice questions (with answers) about electrical
-  engineering.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_electrical_engineering
+"dataset_name": "electrical_engineering"
+"description": "The following are multiple choice questions (with answers) about electrical\
+  \ engineering.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_electrical_engineering"

lm_eval/tasks/mmlu/flan_cot_zeroshot/mmlu_elementary_mathematics.yaml

-dataset_name: elementary_mathematics
-description: 'The following are multiple choice questions (with answers) about elementary
-  mathematics.
-
-
-  '
-include: _mmlu_flan_generative_template_yaml
-task: mmlu_flan_cot_zeroshot_elementary_mathematics
+"dataset_name": "elementary_mathematics"
+"description": "The following are multiple choice questions (with answers) about elementary\
+  \ mathematics.\n\n"
+"group": "mmlu_flan_cot_zeroshot_stem"
+"include": "_mmlu_flan_cot_zeroshot_template_yaml"
+"task": "mmlu_flan_cot_zeroshot_elementary_mathematics"