Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
gaoqiong
lm-evaluation-harness
Commits
96336194
Commit
96336194
authored
May 06, 2024
by
lintangsutawika
Browse files
modify mmlu tasks
parent
0579b304
Changes
59
Hide whitespace changes
Inline
Side-by-side
Showing
20 changed files
with
45 additions
and
51 deletions
+45
-51
lm_eval/tasks/mmlu/default/_mmlu_stem.yaml
lm_eval/tasks/mmlu/default/_mmlu_stem.yaml
+24
-0
lm_eval/tasks/mmlu/default/mmlu.yaml
lm_eval/tasks/mmlu/default/mmlu.yaml
+6
-0
lm_eval/tasks/mmlu/default/mmlu_abstract_algebra.yaml
lm_eval/tasks/mmlu/default/mmlu_abstract_algebra.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_anatomy.yaml
lm_eval/tasks/mmlu/default/mmlu_anatomy.yaml
+0
-2
lm_eval/tasks/mmlu/default/mmlu_astronomy.yaml
lm_eval/tasks/mmlu/default/mmlu_astronomy.yaml
+0
-2
lm_eval/tasks/mmlu/default/mmlu_business_ethics.yaml
lm_eval/tasks/mmlu/default/mmlu_business_ethics.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_clinical_knowledge.yaml
lm_eval/tasks/mmlu/default/mmlu_clinical_knowledge.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_college_biology.yaml
lm_eval/tasks/mmlu/default/mmlu_college_biology.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_college_chemistry.yaml
lm_eval/tasks/mmlu/default/mmlu_college_chemistry.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_college_computer_science.yaml
...val/tasks/mmlu/default/mmlu_college_computer_science.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_college_mathematics.yaml
lm_eval/tasks/mmlu/default/mmlu_college_mathematics.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_college_medicine.yaml
lm_eval/tasks/mmlu/default/mmlu_college_medicine.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_college_physics.yaml
lm_eval/tasks/mmlu/default/mmlu_college_physics.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_computer_security.yaml
lm_eval/tasks/mmlu/default/mmlu_computer_security.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_conceptual_physics.yaml
lm_eval/tasks/mmlu/default/mmlu_conceptual_physics.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_econometrics.yaml
lm_eval/tasks/mmlu/default/mmlu_econometrics.yaml
+0
-2
lm_eval/tasks/mmlu/default/mmlu_electrical_engineering.yaml
lm_eval/tasks/mmlu/default/mmlu_electrical_engineering.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_elementary_mathematics.yaml
lm_eval/tasks/mmlu/default/mmlu_elementary_mathematics.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_formal_logic.yaml
lm_eval/tasks/mmlu/default/mmlu_formal_logic.yaml
+1
-3
lm_eval/tasks/mmlu/default/mmlu_global_facts.yaml
lm_eval/tasks/mmlu/default/mmlu_global_facts.yaml
+1
-3
No files found.
lm_eval/tasks/mmlu/default/_mmlu_stem.yaml
0 → 100644
View file @
96336194
group
:
mmlu_stem
task
:
-
abstract_algebra
-
anatomy
-
astronomy
-
college_biology
-
college_chemistry
-
college_computer_science
-
college_mathematics
-
college_physics
-
computer_security
-
conceptual_physics
-
electrical_engineering
-
elementary_mathematics
-
high_school_biology
-
high_school_chemistry
-
high_school_computer_science
-
high_school_mathematics
-
high_school_physics
-
high_school_statistics
-
machine_learning
group_config
:
aggregate_metric
:
True
weight_by_size
:
True
lm_eval/tasks/mmlu/default/mmlu.yaml
0 → 100644
View file @
96336194
group
:
mmlu
task
:
-
mmlu_stem
-
mmlu_other
-
mmlu_social_sciences
-
mmlu_humanities
lm_eval/tasks/mmlu/default/mmlu_abstract_algebra.yaml
View file @
96336194
"
dataset_name"
:
"
abstract_algebra"
"
dataset_name"
:
"
abstract_algebra"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
abstract
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
abstract
\
\
algebra.
\n\n
"
\
algebra.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_abstract_algebra"
"
task"
:
"
mmlu_abstract_algebra"
"
task_alias"
:
"
abstract
_
algebra"
"
task_alias"
:
"
abstract
algebra"
lm_eval/tasks/mmlu/default/mmlu_anatomy.yaml
View file @
96336194
"
dataset_name"
:
"
anatomy"
"
dataset_name"
:
"
anatomy"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
anatomy.
\n\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
anatomy.
\n\
\n
"
\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_anatomy"
"
task"
:
"
mmlu_anatomy"
"
task_alias"
:
"
anatomy"
"
task_alias"
:
"
anatomy"
lm_eval/tasks/mmlu/default/mmlu_astronomy.yaml
View file @
96336194
"
dataset_name"
:
"
astronomy"
"
dataset_name"
:
"
astronomy"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
astronomy.
\n\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
astronomy.
\n\
\n
"
\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_astronomy"
"
task"
:
"
mmlu_astronomy"
"
task_alias"
:
"
astronomy"
"
task_alias"
:
"
astronomy"
lm_eval/tasks/mmlu/default/mmlu_business_ethics.yaml
View file @
96336194
"
dataset_name"
:
"
business_ethics"
"
dataset_name"
:
"
business_ethics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
business
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
business
\
\
ethics.
\n\n
"
\
ethics.
\n\n
"
"
group"
:
"
mmlu_other"
"
group_alias"
:
"
other"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_business_ethics"
"
task"
:
"
mmlu_business_ethics"
"
task_alias"
:
"
business
_
ethics"
"
task_alias"
:
"
business
ethics"
lm_eval/tasks/mmlu/default/mmlu_clinical_knowledge.yaml
View file @
96336194
"
dataset_name"
:
"
clinical_knowledge"
"
dataset_name"
:
"
clinical_knowledge"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
clinical
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
clinical
\
\
knowledge.
\n\n
"
\
knowledge.
\n\n
"
"
group"
:
"
mmlu_other"
"
group_alias"
:
"
other"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_clinical_knowledge"
"
task"
:
"
mmlu_clinical_knowledge"
"
task_alias"
:
"
clinical
_
knowledge"
"
task_alias"
:
"
clinical
knowledge"
lm_eval/tasks/mmlu/default/mmlu_college_biology.yaml
View file @
96336194
"
dataset_name"
:
"
college_biology"
"
dataset_name"
:
"
college_biology"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
biology.
\n\n
"
\
biology.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_biology"
"
task"
:
"
mmlu_college_biology"
"
task_alias"
:
"
college
_
biology"
"
task_alias"
:
"
college
biology"
lm_eval/tasks/mmlu/default/mmlu_college_chemistry.yaml
View file @
96336194
"
dataset_name"
:
"
college_chemistry"
"
dataset_name"
:
"
college_chemistry"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
chemistry.
\n\n
"
\
chemistry.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_chemistry"
"
task"
:
"
mmlu_college_chemistry"
"
task_alias"
:
"
college
_
chemistry"
"
task_alias"
:
"
college
chemistry"
lm_eval/tasks/mmlu/default/mmlu_college_computer_science.yaml
View file @
96336194
"
dataset_name"
:
"
college_computer_science"
"
dataset_name"
:
"
college_computer_science"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
computer
science.
\n\n
"
\
computer
science.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_computer_science"
"
task"
:
"
mmlu_college_computer_science"
"
task_alias"
:
"
college
_
computer
_
science"
"
task_alias"
:
"
college
computer
science"
lm_eval/tasks/mmlu/default/mmlu_college_mathematics.yaml
View file @
96336194
"
dataset_name"
:
"
college_mathematics"
"
dataset_name"
:
"
college_mathematics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
mathematics.
\n\n
"
\
mathematics.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_mathematics"
"
task"
:
"
mmlu_college_mathematics"
"
task_alias"
:
"
college
_
mathematics"
"
task_alias"
:
"
college
mathematics"
lm_eval/tasks/mmlu/default/mmlu_college_medicine.yaml
View file @
96336194
"
dataset_name"
:
"
college_medicine"
"
dataset_name"
:
"
college_medicine"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
medicine.
\n\n
"
\
medicine.
\n\n
"
"
group"
:
"
mmlu_other"
"
group_alias"
:
"
other"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_medicine"
"
task"
:
"
mmlu_college_medicine"
"
task_alias"
:
"
college
_
medicine"
"
task_alias"
:
"
college
medicine"
lm_eval/tasks/mmlu/default/mmlu_college_physics.yaml
View file @
96336194
"
dataset_name"
:
"
college_physics"
"
dataset_name"
:
"
college_physics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
physics.
\n\n
"
\
physics.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_physics"
"
task"
:
"
mmlu_college_physics"
"
task_alias"
:
"
college
_
physics"
"
task_alias"
:
"
college
physics"
lm_eval/tasks/mmlu/default/mmlu_computer_security.yaml
View file @
96336194
"
dataset_name"
:
"
computer_security"
"
dataset_name"
:
"
computer_security"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
computer
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
computer
\
\
security.
\n\n
"
\
security.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_computer_security"
"
task"
:
"
mmlu_computer_security"
"
task_alias"
:
"
computer
_
security"
"
task_alias"
:
"
computer
security"
lm_eval/tasks/mmlu/default/mmlu_conceptual_physics.yaml
View file @
96336194
"
dataset_name"
:
"
conceptual_physics"
"
dataset_name"
:
"
conceptual_physics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
conceptual
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
conceptual
\
\
physics.
\n\n
"
\
physics.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_conceptual_physics"
"
task"
:
"
mmlu_conceptual_physics"
"
task_alias"
:
"
conceptual
_
physics"
"
task_alias"
:
"
conceptual
physics"
lm_eval/tasks/mmlu/default/mmlu_econometrics.yaml
View file @
96336194
"
dataset_name"
:
"
econometrics"
"
dataset_name"
:
"
econometrics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
econometrics.
\n\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
econometrics.
\n\
\n
"
\n
"
"
group"
:
"
mmlu_social_sciences"
"
group_alias"
:
"
social_sciences"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_econometrics"
"
task"
:
"
mmlu_econometrics"
"
task_alias"
:
"
econometrics"
"
task_alias"
:
"
econometrics"
lm_eval/tasks/mmlu/default/mmlu_electrical_engineering.yaml
View file @
96336194
"
dataset_name"
:
"
electrical_engineering"
"
dataset_name"
:
"
electrical_engineering"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
electrical
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
electrical
\
\
engineering.
\n\n
"
\
engineering.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_electrical_engineering"
"
task"
:
"
mmlu_electrical_engineering"
"
task_alias"
:
"
electrical
_
engineering"
"
task_alias"
:
"
electrical
engineering"
lm_eval/tasks/mmlu/default/mmlu_elementary_mathematics.yaml
View file @
96336194
"
dataset_name"
:
"
elementary_mathematics"
"
dataset_name"
:
"
elementary_mathematics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
elementary
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
elementary
\
\
mathematics.
\n\n
"
\
mathematics.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_elementary_mathematics"
"
task"
:
"
mmlu_elementary_mathematics"
"
task_alias"
:
"
elementary
_
mathematics"
"
task_alias"
:
"
elementary
mathematics"
lm_eval/tasks/mmlu/default/mmlu_formal_logic.yaml
View file @
96336194
"
dataset_name"
:
"
formal_logic"
"
dataset_name"
:
"
formal_logic"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
formal
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
formal
\
\
logic.
\n\n
"
\
logic.
\n\n
"
"
group"
:
"
mmlu_humanities"
"
group_alias"
:
"
humanities"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_formal_logic"
"
task"
:
"
mmlu_formal_logic"
"
task_alias"
:
"
formal
_
logic"
"
task_alias"
:
"
formal
logic"
lm_eval/tasks/mmlu/default/mmlu_global_facts.yaml
View file @
96336194
"
dataset_name"
:
"
global_facts"
"
dataset_name"
:
"
global_facts"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
global
\
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
global
\
\
facts.
\n\n
"
\
facts.
\n\n
"
"
group"
:
"
mmlu_other"
"
group_alias"
:
"
other"
"
include"
:
"
_default_template_yaml"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_global_facts"
"
task"
:
"
mmlu_global_facts"
"
task_alias"
:
"
global
_
facts"
"
task_alias"
:
"
global
facts"
Prev
1
2
3
Next
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment