add arc_challenge_mt (#1900)

* add arc_challenge_mt

* add README

* add icelandic

lm_eval/tasks/arc_mt/README.md

+# arc mt
+
+arc mt is an implementation of tasks to support machine translated arc
+challenge evals, to improve eval support across a number of additional
+languages.
+
+The main page for the effort is
+[here](https://huggingface.co/datasets/LumiOpen/arc_challenge_mt) and we will
+include more data and analysis there.
+
+Initial datasets include a number of European languages, and we plan to expand
+more in the future.

lm_eval/tasks/arc_mt/arc_challenge_mt_da.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_da
+dataset_name: da

lm_eval/tasks/arc_mt/arc_challenge_mt_de.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_de
+dataset_name: de

lm_eval/tasks/arc_mt/arc_challenge_mt_el.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_el
+dataset_name: el

lm_eval/tasks/arc_mt/arc_challenge_mt_es.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_es
+dataset_name: es

lm_eval/tasks/arc_mt/arc_challenge_mt_fi.yaml

+group:
+  - arc_challenge_mt
+task: arc_challenge_mt_fi
+dataset_path: LumiOpen/arc_challenge_mt
+dataset_name: fi
+output_type: multiple_choice
+training_split: train
+validation_split: validation
+test_split: test
+doc_to_text: "Question: {{question}}\nAnswer:"
+doc_to_target: "{{choices.label.index(answerKey)}}"
+doc_to_choice: "{{choices.text}}"
+should_decontaminate: true
+doc_to_decontamination_query: "Question: {{question}}\nAnswer:"
+metric_list:
+  - metric: acc
+    aggregation: mean
+    higher_is_better: true
+  - metric: acc_norm
+    aggregation: mean
+    higher_is_better: true
+metadata:
+  version: 1.0

lm_eval/tasks/arc_mt/arc_challenge_mt_hu.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_hu
+dataset_name: hu

lm_eval/tasks/arc_mt/arc_challenge_mt_is.yaml

+group:
+  - arc_challenge_mt
+task: arc_challenge_mt_is
+dataset_path: mideind/icelandic-arc-challenge
+output_type: multiple_choice
+training_split: train
+validation_split: validation
+test_split: test
+doc_to_text: "Question: {{question}}\nAnswer:"
+doc_to_target: "{{choices.label.index(answerKey)}}"
+doc_to_choice: "{{choices.text}}"
+should_decontaminate: true
+doc_to_decontamination_query: "Question: {{question}}\nAnswer:"
+metric_list:
+  - metric: acc
+    aggregation: mean
+    higher_is_better: true
+  - metric: acc_norm
+    aggregation: mean
+    higher_is_better: true
+metadata:
+  version: 1.0

lm_eval/tasks/arc_mt/arc_challenge_mt_it.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_it
+dataset_name: it

lm_eval/tasks/arc_mt/arc_challenge_mt_nb.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_nb
+dataset_name: nb

lm_eval/tasks/arc_mt/arc_challenge_mt_pl.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_pl
+dataset_name: pl

lm_eval/tasks/arc_mt/arc_challenge_mt_pt.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_pt
+dataset_name: pt

lm_eval/tasks/arc_mt/arc_challenge_mt_sv.yaml

+include: arc_challenge_mt_fi.yaml
+task: arc_challenge_mt_sv
+dataset_name: sv