resolved conflict

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_05/b.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_05
+task: arc_challenge_alt_ov_05b
+doc_to_text: !function ../styles.template_05
+doc_to_choice: !function ../styles.choice_05b
+doc_to_decontamination_query: !function ../styles.template_05

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_05/c.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_05
+task: arc_challenge_alt_ov_05c
+doc_to_text: !function ../styles.template_05
+doc_to_choice: !function ../styles.choice_05c
+doc_to_decontamination_query: !function ../styles.template_05

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_06/a.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_06
+task: arc_challenge_alt_ov_06a
+doc_to_text: !function ../styles.template_06
+doc_to_choice: !function ../styles.choice_06a
+doc_to_decontamination_query: !function ../styles.template_06

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_06/b.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_06
+task: arc_challenge_alt_ov_06b
+doc_to_text: !function ../styles.template_06
+doc_to_choice: !function ../styles.choice_06b
+doc_to_decontamination_query: !function ../styles.template_06

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_06/c.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_06
+task: arc_challenge_alt_ov_06c
+doc_to_text: !function ../styles.template_06
+doc_to_choice: !function ../styles.choice_06c
+doc_to_decontamination_query: !function ../styles.template_06

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_07/a.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_07
+task: arc_challenge_alt_ov_07a
+doc_to_text: !function ../styles.template_07
+doc_to_choice: !function ../styles.choice_07a
+doc_to_decontamination_query: !function ../styles.template_07

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_07/b.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_07
+task: arc_challenge_alt_ov_07b
+doc_to_text: !function ../styles.template_07
+doc_to_choice: !function ../styles.choice_07b
+doc_to_decontamination_query: !function ../styles.template_07

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_07/c.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_07
+task: arc_challenge_alt_ov_07c
+doc_to_text: !function ../styles.template_07
+doc_to_choice: !function ../styles.choice_07c
+doc_to_decontamination_query: !function ../styles.template_07

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_08/a.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_08
+task: arc_challenge_alt_ov_08a
+doc_to_text: !function ../styles.template_08
+doc_to_choice: !function ../styles.choice_08a
+doc_to_decontamination_query: !function ../styles.template_08

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_08/b.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_08
+task: arc_challenge_alt_ov_08b
+doc_to_text: !function ../styles.template_08
+doc_to_choice: !function ../styles.choice_08b
+doc_to_decontamination_query: !function ../styles.template_08

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/style_08/c.yaml

+include: ../_arc_challenge_alt_yaml
+group: arc_challenge_alt_ov_08
+task: arc_challenge_alt_ov_08c
+doc_to_text: !function ../styles.template_08
+doc_to_choice: !function ../styles.choice_08c
+doc_to_decontamination_query: !function ../styles.template_08

lm_eval/tasks/arc/alternative_worlds/arc_challenge/output_variation/styles.py

+import string
+from functools import partial
+
+
+def doc_to_text_base(alphabet, style, doc):
+
+    choices = doc["choices"]["text"]
+    num = len(choices)
+
+    letter_list = [style.format(letter) for letter in alphabet[0:num]]
+
+    if "\t" in style:
+        choice_string = "{}{}"
+    else:
+        choice_string = "{} {}"
+
+    doc_to_text = "\n".join(
+        [
+            "Question: " + doc["question"],
+        ]
+        + [choice_string.format(i, j) for i, j in zip(letter_list, choices)]
+        + ["Answer:"]
+    )
+
+    return doc_to_text
+
+
+# Full continuation
+def choice_A(doc):
+    return doc["choices"]["text"]
+
+
+# Letters only
+def choice_B(alphabet, style, doc):
+
+    choices = doc["choices"]["text"]
+    num = len(choices)
+
+    letter_list = [style.format(letter) for letter in alphabet[0:num]]
+    if "\t" in style:
+        letter_list = [letter.replace("\t", "") for letter in letter_list]
+
+    return letter_list
+
+
+# Letters + Full continuation
+def choice_C(alphabet, style, doc):
+
+    choices = doc["choices"]["text"]
+    num = len(choices)
+
+    letter_list = [style.format(letter) for letter in alphabet[0:num]]
+    if "\t" not in style:
+        letter_list = [letter + " " for letter in letter_list]
+
+    return [letter + choice for letter, choice in zip(letter_list, choices)]
+
+
+template_01 = partial(doc_to_text_base, string.ascii_lowercase, "({})")
+choice_01a = choice_A
+choice_01b = partial(choice_B, string.ascii_lowercase, "({})")
+choice_01c = partial(choice_C, string.ascii_lowercase, "({})")
+template_02 = partial(doc_to_text_base, string.ascii_lowercase, "{})")
+choice_02a = choice_A
+choice_02b = partial(choice_B, string.ascii_lowercase, "{})")
+choice_02c = partial(choice_C, string.ascii_lowercase, "{})")
+template_03 = partial(doc_to_text_base, string.ascii_lowercase, "{}.")
+choice_03a = choice_A
+choice_03b = partial(choice_B, string.ascii_lowercase, "{}.")
+choice_03c = partial(choice_C, string.ascii_lowercase, "{}.")
+template_04 = partial(doc_to_text_base, string.ascii_lowercase, "{}\t")
+choice_04a = choice_A
+choice_04b = partial(choice_B, string.ascii_lowercase, "{}\t")
+choice_04c = partial(choice_C, string.ascii_lowercase, "{}\t")
+template_05 = partial(doc_to_text_base, string.ascii_uppercase, "({})")
+choice_05a = choice_A
+choice_05b = partial(choice_B, string.ascii_uppercase, "({})")
+choice_05c = partial(choice_C, string.ascii_uppercase, "({})")
+template_06 = partial(doc_to_text_base, string.ascii_uppercase, "{})")
+choice_06a = choice_A
+choice_06b = partial(choice_B, string.ascii_uppercase, "{})")
+choice_06c = partial(choice_C, string.ascii_uppercase, "{})")
+template_07 = partial(doc_to_text_base, string.ascii_uppercase, "{}.")
+choice_07a = choice_A
+choice_07b = partial(choice_B, string.ascii_uppercase, "{}.")
+choice_07c = partial(choice_C, string.ascii_uppercase, "{}.")
+template_08 = partial(doc_to_text_base, string.ascii_uppercase, "{}\t")
+choice_08a = choice_A
+choice_08b = partial(choice_B, string.ascii_uppercase, "{}\t")
+choice_08c = partial(choice_C, string.ascii_uppercase, "{}\t")

lm_eval/tasks/arc/alternative_worlds/arc_challenge/prompt_variation/_arc_challenge_alt_yaml

+dataset_path: ai2_arc
+dataset_name: 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
+  - metric: brier_score
+    aggregation: brier_score
+    higher_is_better: false

lm_eval/tasks/arc/alternative_worlds/arc_challenge/prompt_variation/style_01.yaml

+include: _arc_challenge_alt_yaml
+group: arc_challenge_alt_pv
+task: arc_challenge_alt_pv_01
+doc_to_text: "{{question}}"
+doc_to_decontamination_query: "{{question}}"

lm_eval/tasks/arc/alternative_worlds/arc_challenge/prompt_variation/style_02.yaml

+include: _arc_challenge_alt_yaml
+group: arc_challenge_alt_pv
+task: arc_challenge_alt_pv_02
+doc_to_text: "Q: {{question}}\nA:"
+doc_to_decontamination_query: "Q: {{question}}\nA:"

lm_eval/tasks/arc/alternative_worlds/arc_challenge/prompt_variation/style_03.yaml

+include: _arc_challenge_alt_yaml
+group: arc_challenge_alt_pv
+task: arc_challenge_alt_pv_03
+doc_to_text: "Question: {{question}}\nAnswer:"
+doc_to_decontamination_query: "Question: {{question}}\nAnswer:"

lm_eval/tasks/arc/alternative_worlds/arc_easy/README.md

+
+
+Investigate affect of letter options
+- (A)
+- A)
+- A.
+- A\t
+- (a)
+- a)
+- a.
+- a\t
+
+Answer types:
+- letters only
+    - original option
+    - just letter
+- letters + continuation
+    - original option
+    - just letter
+- continuation

lm_eval/tasks/arc/alternative_worlds/arc_easy/output_variation/_arc_easy_alt_yaml

+group:
+  - ai2_arc
+task: arc_easy
+dataset_path: ai2_arc
+dataset_name: ARC-Easy
+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
+  - metric: brier_score
+    aggregation: brier_score
+    higher_is_better: false

lm_eval/tasks/arc/alternative_worlds/arc_easy/output_variation/arc_easy_alt.yaml

+group: arc_easy_alt_ov
+task:
+  - arc_easy_alt_ov_01
+  - arc_easy_alt_ov_02
+  - arc_easy_alt_ov_03
+  - arc_easy_alt_ov_04
+  - arc_easy_alt_ov_05
+  - arc_easy_alt_ov_06
+  - arc_easy_alt_ov_07
+  - arc_easy_alt_ov_08

lm_eval/tasks/arc/alternative_worlds/arc_easy/output_variation/style_01/a.yaml

+include: ../_arc_easy_alt_yaml
+group: arc_easy_alt_ov_01
+task: arc_easy_alt_ov_01a
+doc_to_text: !function ../styles.template_01
+doc_to_choice: !function ../styles.choice_01a
+doc_to_decontamination_query: !function ../styles.template_01