Merge branch 'main' of https://github.com/EleutherAI/lm-evaluation-harness into mela

lm_eval/tasks/agieval/gaokao-physics.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_cn
 task: agieval_gaokao_physics
 dataset_path: hails/agieval-gaokao-physics

lm_eval/tasks/agieval/jec-qa-ca.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_cn
 task: agieval_jec_qa_ca
 dataset_path: hails/agieval-jec-qa-ca

lm_eval/tasks/agieval/jec-qa-kd.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_cn
 task: agieval_jec_qa_kd
 dataset_path: hails/agieval-jec-qa-kd

lm_eval/tasks/agieval/logiqa-en.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_nous
-  - agieval_en
 task: agieval_logiqa_en
 dataset_path: hails/agieval-logiqa-en

lm_eval/tasks/agieval/logiqa-zh.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_cn
 task: agieval_logiqa_zh
 dataset_path: hails/agieval-logiqa-zh

lm_eval/tasks/agieval/lsat-ar.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_nous
-  - agieval_en
 task: agieval_lsat_ar
 dataset_path: hails/agieval-lsat-ar

lm_eval/tasks/agieval/lsat-lr.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_nous
-  - agieval_en
 task: agieval_lsat_lr
 dataset_path: hails/agieval-lsat-lr

lm_eval/tasks/agieval/lsat-rc.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_nous
-  - agieval_en
 task: agieval_lsat_rc
 dataset_path: hails/agieval-lsat-rc

lm_eval/tasks/agieval/math.yaml

-group:
-  - agieval
-  - agieval_en
 task: agieval_math
 dataset_path: hails/agieval-math
 dataset_name: null

lm_eval/tasks/agieval/sat-en-without-passage.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_nous
-  - agieval_en
 task: agieval_sat_en_without_passage
 dataset_path: hails/agieval-sat-en-without-passage

lm_eval/tasks/agieval/sat-en.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_nous
-  - agieval_en
 task: agieval_sat_en
 dataset_path: hails/agieval-sat-en

lm_eval/tasks/agieval/sat-math.yaml

 include: aqua-rat.yaml
-group:
-  - agieval
-  - agieval_nous
-  - agieval_en
 task: agieval_sat_math
 dataset_path: hails/agieval-sat-math

lm_eval/tasks/anli/anli_r1.yaml

-group:
+tag:
   - anli
 task: anli_r1
 dataset_path: anli

lm_eval/tasks/arabicmmlu/README.md

+# ArabicMMLU
+
+### Paper
+
+Title: ArabicMMLU: Assessing Massive Multitask Language Understanding in Arabic
+
+Abstract: https://arxiv.org/abs/2402.12840
+
+The focus of language model evaluation has
+transitioned towards reasoning and knowledge intensive tasks, driven by advancements in pretraining large models. While state-of-the-art models are partially trained on large Arabic texts, evaluating their performance in Arabic remains challenging due to the limited availability of relevant datasets. To bridge this gap, we present ArabicMMLU, the first multi-task language understanding benchmark for Arabic language, sourced from school exams across diverse educational levels in different countries spanning North Africa, the Levant, and the Gulf regions. Our data comprises 40 tasks and 14,575 multiple-choice questions in Modern Standard Arabic (MSA), and is carefully constructed by collaborating with native speakers in the region. Our comprehensive evaluations of 35 models reveal substantial room for improvement, particularly among the best open-source models. Notably, BLOOMZ, mT0, LLama2, and Falcon struggle to achieve a score of 50%, while even the top-performing Arabic centric model only achieves a score of 62.3%.
+
+The authors of the paper conducted studies by varying the language of the initial prompt and answer keys between English and Arabic. However, they set English initial prompts and answer keys as the standard, which is the version implemented in this task.
+
+Homepage: https://github.com/mbzuai-nlp/ArabicMMLU
+
+
+### Citation
+
+```
+@misc{koto2024arabicmmlu,
+      title={ArabicMMLU: Assessing Massive Multitask Language Understanding in Arabic},
+      author={Fajri Koto and Haonan Li and Sara Shatnawi and Jad Doughman and Abdelrahman Boda Sadallah and Aisha Alraeesi and Khalid Almubarak and Zaid Alyafeai and Neha Sengupta and Shady Shehata and Nizar Habash and Preslav Nakov and Timothy Baldwin},
+      year={2024},
+      eprint={2402.12840},
+      archivePrefix={arXiv},
+      primaryClass={id='cs.CL' full_name='Computation and Language' is_active=True alt_name='cmp-lg' in_archive='cs' is_general=False description='Covers natural language processing. Roughly includes material in ACM Subject Class I.2.7. Note that work on artificial languages (programming languages, logics, formal systems) that does not explicitly address natural-language issues broadly construed (natural-language processing, computational linguistics, speech, text retrieval, etc.) is not appropriate for this area.'}
+}
+```
+
+### Groups and Tasks
+
+#### Groups
+
+* `arabicmmlu`: evaluates all ArabicMMLU tasks.
+
+* `arabicmmlu_stem`: evaluates STEM ArabicMMLU tasks.
+* `arabicmmlu_stem_social_science`: evaluates social science ArabicMMLU tasks.
+* `arabicmmlu_stem_humanities`: evaluates humanities ArabicMMLU tasks.
+* `arabicmmlu_stem_language`: evaluates Arabic language ArabicMMLU tasks.
+* `arabicmmlu_stem_other`: evaluates other ArabicMMLU tasks.

lm_eval/tasks/arabicmmlu/_arabicmmlu.yaml

+group: arabicmmlu
+task:
+- arabicmmlu_other
+- arabicmmlu_social_science
+- arabicmmlu_humanities
+- arabicmmlu_stem
+- arabicmmlu_language
+aggregate_metric_list:
+  - metric: acc
+    weight_by_size: True
+metadata:
+  version: 0

lm_eval/tasks/arabicmmlu/_arabicmmlu_humanities.yaml

+group: arabicmmlu_humanities
+group_alias: Humanities
+task:
+  - arabicmmlu_humanities_tasks
+aggregate_metric_list:
+  - metric: acc
+    weight_by_size: True
+metadata:
+  version: 0

lm_eval/tasks/arabicmmlu/_arabicmmlu_language.yaml

+group: arabicmmlu_language
+group_alias: Language
+task:
+  - arabicmmlu_language_tasks
+aggregate_metric_list:
+  - metric: acc
+    weight_by_size: True
+metadata:
+  version: 0

lm_eval/tasks/arabicmmlu/_arabicmmlu_other.yaml

+group: arabicmmlu_other
+group_alias: Other
+task:
+  - arabicmmlu_other_tasks
+aggregate_metric_list:
+  - metric: acc
+    weight_by_size: True
+metadata:
+  version: 0

lm_eval/tasks/arabicmmlu/_arabicmmlu_social_science.yaml

+group: arabicmmlu_social_science
+group_alias: Social Science
+task:
+  - arabicmmlu_social_science_tasks
+aggregate_metric_list:
+  - metric: acc
+    weight_by_size: True
+metadata:
+  version: 0

lm_eval/tasks/arabicmmlu/_arabicmmlu_stem.yaml

+group: arabicmmlu_stem
+group_alias: STEM
+task:
+  - arabicmmlu_stem_tasks
+aggregate_metric_list:
+  - metric: acc
+    weight_by_size: True
+metadata:
+  version: 0