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
104292ff
Commit
104292ff
authored
May 16, 2024
by
lintangsutawika
Browse files
update mmlu
parent
923f3e81
Changes
62
Hide whitespace changes
Inline
Side-by-side
Showing
20 changed files
with
47 additions
and
54 deletions
+47
-54
lm_eval/tasks/mmlu/default/_mmlu.yaml
lm_eval/tasks/mmlu/default/_mmlu.yaml
+4
-24
lm_eval/tasks/mmlu/default/_mmlu_humanities.yaml
lm_eval/tasks/mmlu/default/_mmlu_humanities.yaml
+7
-0
lm_eval/tasks/mmlu/default/_mmlu_other.yaml
lm_eval/tasks/mmlu/default/_mmlu_other.yaml
+7
-0
lm_eval/tasks/mmlu/default/_mmlu_social_sciences.yaml
lm_eval/tasks/mmlu/default/_mmlu_social_sciences.yaml
+7
-0
lm_eval/tasks/mmlu/default/_mmlu_stem.yaml
lm_eval/tasks/mmlu/default/_mmlu_stem.yaml
+7
-0
lm_eval/tasks/mmlu/default/mmlu_abstract_algebra.yaml
lm_eval/tasks/mmlu/default/mmlu_abstract_algebra.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_anatomy.yaml
lm_eval/tasks/mmlu/default/mmlu_anatomy.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_astronomy.yaml
lm_eval/tasks/mmlu/default/mmlu_astronomy.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_business_ethics.yaml
lm_eval/tasks/mmlu/default/mmlu_business_ethics.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_clinical_knowledge.yaml
lm_eval/tasks/mmlu/default/mmlu_clinical_knowledge.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_college_biology.yaml
lm_eval/tasks/mmlu/default/mmlu_college_biology.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_college_chemistry.yaml
lm_eval/tasks/mmlu/default/mmlu_college_chemistry.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_college_computer_science.yaml
...val/tasks/mmlu/default/mmlu_college_computer_science.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_college_mathematics.yaml
lm_eval/tasks/mmlu/default/mmlu_college_mathematics.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_college_medicine.yaml
lm_eval/tasks/mmlu/default/mmlu_college_medicine.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_college_physics.yaml
lm_eval/tasks/mmlu/default/mmlu_college_physics.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_computer_security.yaml
lm_eval/tasks/mmlu/default/mmlu_computer_security.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_conceptual_physics.yaml
lm_eval/tasks/mmlu/default/mmlu_conceptual_physics.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_econometrics.yaml
lm_eval/tasks/mmlu/default/mmlu_econometrics.yaml
+1
-2
lm_eval/tasks/mmlu/default/mmlu_electrical_engineering.yaml
lm_eval/tasks/mmlu/default/mmlu_electrical_engineering.yaml
+1
-2
No files found.
lm_eval/tasks/mmlu/default/_mmlu.yaml
View file @
104292ff
group
:
mmlu
task
:
-
group
:
stem
task
:
-
mmlu_stem
aggregate_metric
:
True
weight_by_size
:
True
tag_to_task
:
True
-
group
:
other
task
:
-
mmlu_other
aggregate_metric
:
True
weight_by_size
:
True
tag_to_task
:
True
-
group
:
social sciences
task
:
-
mmlu_social_sciences
aggregate_metric
:
True
weight_by_size
:
True
tag_to_task
:
True
-
group
:
humanities
task
:
-
mmlu_humanities
aggregate_metric
:
True
weight_by_size
:
True
tag_to_task
:
True
-
mmlu_stem
-
mmlu_other
-
mmlu_social_sciences
-
mmlu_humanities
aggregate_metric
:
True
weight_by_size
:
True
version
:
1
lm_eval/tasks/mmlu/default/_mmlu_humanities.yaml
0 → 100644
View file @
104292ff
group
:
mmlu_humanities
group_alias
:
humanities
task
:
-
mmlu_humanities_tasks
aggregate_metric
:
True
weight_by_size
:
True
version
:
1
lm_eval/tasks/mmlu/default/_mmlu_other.yaml
0 → 100644
View file @
104292ff
group
:
mmlu_other
group_alias
:
other
task
:
-
mmlu_other_tasks
aggregate_metric
:
True
weight_by_size
:
True
version
:
1
lm_eval/tasks/mmlu/default/_mmlu_social_sciences.yaml
0 → 100644
View file @
104292ff
group
:
mmlu_social_sciences
group_alias
:
social sciences
task
:
-
mmlu_social_sciences_tasks
aggregate_metric
:
True
weight_by_size
:
True
version
:
1
lm_eval/tasks/mmlu/default/_mmlu_stem.yaml
0 → 100644
View file @
104292ff
group
:
mmlu_stem
group_alias
:
stem
task
:
-
mmlu_stem_tasks
aggregate_metric
:
True
weight_by_size
:
True
version
:
1
lm_eval/tasks/mmlu/default/mmlu_abstract_algebra.yaml
View file @
104292ff
"
dataset_name"
:
"
abstract_algebra"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
abstract
\
\
algebra.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_abstract_algebra"
"
task_alias"
:
"
abstract_algebra"
lm_eval/tasks/mmlu/default/mmlu_anatomy.yaml
View file @
104292ff
"
dataset_name"
:
"
anatomy"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
anatomy.
\n\
\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_anatomy"
"
task_alias"
:
"
anatomy"
lm_eval/tasks/mmlu/default/mmlu_astronomy.yaml
View file @
104292ff
"
dataset_name"
:
"
astronomy"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
astronomy.
\n\
\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_astronomy"
"
task_alias"
:
"
astronomy"
lm_eval/tasks/mmlu/default/mmlu_business_ethics.yaml
View file @
104292ff
"
dataset_name"
:
"
business_ethics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
business
\
\
ethics.
\n\n
"
"
group"
:
"
mmlu_other"
"
group_alias"
:
"
other"
"
group"
:
"
mmlu_other_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_business_ethics"
"
task_alias"
:
"
business_ethics"
lm_eval/tasks/mmlu/default/mmlu_clinical_knowledge.yaml
View file @
104292ff
"
dataset_name"
:
"
clinical_knowledge"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
clinical
\
\
knowledge.
\n\n
"
"
group"
:
"
mmlu_other"
"
group_alias"
:
"
other"
"
group"
:
"
mmlu_other_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_clinical_knowledge"
"
task_alias"
:
"
clinical_knowledge"
lm_eval/tasks/mmlu/default/mmlu_college_biology.yaml
View file @
104292ff
"
dataset_name"
:
"
college_biology"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
biology.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_biology"
"
task_alias"
:
"
college_biology"
lm_eval/tasks/mmlu/default/mmlu_college_chemistry.yaml
View file @
104292ff
"
dataset_name"
:
"
college_chemistry"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
chemistry.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_chemistry"
"
task_alias"
:
"
college_chemistry"
lm_eval/tasks/mmlu/default/mmlu_college_computer_science.yaml
View file @
104292ff
"
dataset_name"
:
"
college_computer_science"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
computer
science.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_computer_science"
"
task_alias"
:
"
college_computer_science"
lm_eval/tasks/mmlu/default/mmlu_college_mathematics.yaml
View file @
104292ff
"
dataset_name"
:
"
college_mathematics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
mathematics.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_mathematics"
"
task_alias"
:
"
college_mathematics"
lm_eval/tasks/mmlu/default/mmlu_college_medicine.yaml
View file @
104292ff
"
dataset_name"
:
"
college_medicine"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
medicine.
\n\n
"
"
group"
:
"
mmlu_other"
"
group_alias"
:
"
other"
"
group"
:
"
mmlu_other_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_medicine"
"
task_alias"
:
"
college_medicine"
lm_eval/tasks/mmlu/default/mmlu_college_physics.yaml
View file @
104292ff
"
dataset_name"
:
"
college_physics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
college
\
\
physics.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_college_physics"
"
task_alias"
:
"
college_physics"
lm_eval/tasks/mmlu/default/mmlu_computer_security.yaml
View file @
104292ff
"
dataset_name"
:
"
computer_security"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
computer
\
\
security.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_computer_security"
"
task_alias"
:
"
computer_security"
lm_eval/tasks/mmlu/default/mmlu_conceptual_physics.yaml
View file @
104292ff
"
dataset_name"
:
"
conceptual_physics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
conceptual
\
\
physics.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_conceptual_physics"
"
task_alias"
:
"
conceptual_physics"
lm_eval/tasks/mmlu/default/mmlu_econometrics.yaml
View file @
104292ff
"
dataset_name"
:
"
econometrics"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
econometrics.
\n\
\n
"
"
group"
:
"
mmlu_social_sciences"
"
group_alias"
:
"
social_sciences"
"
group"
:
"
mmlu_social_sciences_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_econometrics"
"
task_alias"
:
"
econometrics"
lm_eval/tasks/mmlu/default/mmlu_electrical_engineering.yaml
View file @
104292ff
"
dataset_name"
:
"
electrical_engineering"
"
description"
:
"
The
following
are
multiple
choice
questions
(with
answers)
about
electrical
\
\
engineering.
\n\n
"
"
group"
:
"
mmlu_stem"
"
group_alias"
:
"
stem"
"
group"
:
"
mmlu_stem_tasks"
"
include"
:
"
_default_template_yaml"
"
task"
:
"
mmlu_electrical_engineering"
"
task_alias"
:
"
electrical_engineering"
Prev
1
2
3
4
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