Fewshot refactor (#3227)

* overhaul `ContextSampler`

* refactor masakhapos

* move multi_target to `exact_match`

* remove doc_to_choice from `boolq-seq2seq`

* remove doc_to_choice in generation process_results

* Remove unused `doc_to_choice` and fix superglue whitespaces

* require multiple_inputs and multiple_targets to be explicitly set in taskconfig

* fix copa; better logging in task init

* fix doc_to_target to return int rather than str (deprecated)

* fix processing regression; recursively parse lists fron template

* remove redundant jinja parsing logic

* remove promptsource

* for multiple_inputs use `doc_to_text: list[str]``

* Refactor `ContextSampler` `fewshot_context`

* fix multiple_input context

* fix `target_delimiter` with `gen_prefix`

* `doc_to_text` is list for multiple_inputs

* Refactor `count_bytes` and `count_words` methods to `@staticmethod`

* make has_*(train/test/validation) to properties

* remove `multi_target` `generate_until`

* `fix doc_to_target/multiple_targets handling add tests

* rename `multi_target` to `multiple_targets`

* evalaute list when multiple targets

* allow doc_to_target to return list

* Remove gen_prefix space and add warning (#3239)

* Remove gen_prefix space and add warning

* fix null gen_prefix bug again

* use git tests

---------
Co-authored-by: Boaz Ben-Dov <bendboaz@gmail.com>

lm_eval/tasks/super_glue/copa/default.yaml

 tag:
   - super-glue-lm-eval-v1
 task: copa
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: copa
 output_type: multiple_choice
 training_split: train
 validation_split: validation
 doc_to_text: !function utils.doc_to_text
-doc_to_target: !function utils.doc_to_target
-doc_to_choice: !function utils.doc_to_choice
+doc_to_target: label
+doc_to_choice: ["{{ choice1 }}",  "{{ choice2 }}"]
 metric_list:
   - metric: acc
 metadata:

lm_eval/tasks/super_glue/copa/t5-prompt.yaml

 tag:
   - super-glue-t5-prompt
 task: super_glue-copa-t5-prompt
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: copa
 training_split: train
 validation_split: validation
 output_type: generate_until
 doc_to_text: "copa choice1: {{choice1}} choice2: {{choice2}} premise: {{premise}} question: {{question}}"
-doc_to_target: label
-doc_to_choice: ['choice1', 'choice2']
+doc_to_target: "{{ [choice1, choice2][label|int] }}"
 generation_kwargs:
   until:
     - "</s>"

lm_eval/tasks/super_glue/multirc/default.yaml

 tag:
   - super-glue-lm-eval-v1
 task: multirc
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: multirc
 output_type: multiple_choice
 training_split: train

lm_eval/tasks/super_glue/multirc/t5-prompt.yaml

 tag:
   - super-glue-t5-prompt
 task: super_glue-multirc-t5-prompt
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: multirc
 training_split: train
 validation_split: validation
 output_type: generate_until
-doc_to_text: "multirc question: {{question}} answer: {{answer}} paragraph: {{paragraph}}"
-doc_to_target: label
-doc_to_choice: "{% set group_id = idx.question|string %}{{[group_id+'_False', group_id+'_True']}}"
+doc_to_text: "multirc question: {{question}} answer: {{answer}} paragraph: {{paragraph}}|trim"
+doc_to_target: "{% set group_id = idx.question|string %}{{[group_id+'_False', group_id+'_True'][label]}}"
 generation_kwargs:
   until:
     - "</s>"

lm_eval/tasks/super_glue/record/default.yaml

 tag:
   - super-glue-lm-eval-v1
 task: record
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: record
 output_type: multiple_choice
 training_split: train
@@ -11,6 +11,7 @@ doc_to_target: !function util.doc_to_target
 doc_to_choice: !function util.doc_to_choice
 process_docs: !function util.process_docs
 process_results: !function util.process_results
+target_delimiter: ""
 metric_list:
   - metric: f1
     aggregation: mean

lm_eval/tasks/super_glue/record/t5-prompt.yaml

 tag:
   - super-glue-t5-prompt
 task: super_glue-record-t5-prompt
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: record
 validation_split: validation
 output_type: generate_until

lm_eval/tasks/super_glue/record/util.py

@@ -19,7 +19,7 @@ def format_answer(query, entity):
 
 def doc_to_target(doc):
     # We only output the first correct entity in a doc
-    return format_answer(query=doc["query"], entity=doc["answers"][0])
+    return doc["entities"].index(doc["answers"][0])
 
 
 def doc_to_choice(doc):

lm_eval/tasks/super_glue/rte/default.yaml

 tag:
   - super-glue-lm-eval-v1
 task: sglue_rte
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: rte
 output_type: multiple_choice
 training_split: train
 validation_split: validation
 doc_to_text: "{{premise}}\nQuestion: {{hypothesis}} True or False?\nAnswer:"
 doc_to_target: label
-doc_to_choice: ['True', 'False']
+doc_to_choice: ["True", "False"]
 metric_list:
   - metric: acc
 metadata:

lm_eval/tasks/super_glue/rte/t5-prompt.yaml

 tag:
   - super-glue-t5-prompt
 task: super_glue-rte-t5-prompt
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: rte
 training_split: train
 validation_split: validation
 output_type: generate_until
 doc_to_text: "rte hypothesis: {{hypothesis}} premise: {{premise}}"
-doc_to_target: label
-doc_to_choice: ['entailment', 'not_entailment']
+doc_to_target: "{{ ['entailment', 'not_entailment'][label|int] }}"
 generation_kwargs:
   until:
     - "</s>"

lm_eval/tasks/super_glue/wic/default.yaml

 tag:
   - super-glue-lm-eval-v1
 task: "wic"
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: wic
 output_type: multiple_choice
 training_split: train
 validation_split: validation
 doc_to_text: "Sentence 1: {{sentence1}}\nSentence 2: {{sentence2}}\nQuestion: Is the word '{{sentence1[start1:end1]}}' used in the same way in the two sentences above?\nAnswer:"
 doc_to_target: label
-doc_to_choice: ['no', 'yes']
+doc_to_choice: ["no", "yes"]
 metric_list:
   - metric: acc
 metadata:

lm_eval/tasks/super_glue/wic/t5-prompt.yaml

 tag:
   - super-glue-t5-prompt
 task: super_glue-wic-t5-prompt
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: wic
 training_split: train
 validation_split: validation
 output_type: generate_until
 doc_to_text: "wic sentence1: {{sentence1}} sentence2: {{sentence2}} word: {{word}}"
-doc_to_target: label
-doc_to_choice: ['False', 'True']
+doc_to_target: "{{ ['False', 'True'][label|int] }}"
 generation_kwargs:
   until:
     - "</s>"

lm_eval/tasks/super_glue/wsc/default.yaml

 tag:
   - super-glue-lm-eval-v1
 task: wsc
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: wsc.fixed
 output_type: multiple_choice
 training_split: train
 validation_split: validation
 doc_to_text: !function preprocess_wsc.default_doc_to_text
 doc_to_target: label
-doc_to_choice: ['no', 'yes']
+doc_to_choice: ["no", "yes"]
 metric_list:
   - metric: acc
 metadata:

lm_eval/tasks/super_glue/wsc/t5-prompt.yaml

 tag:
   - super-glue-t5-prompt
 task: super_glue-wsc-t5-prompt
-dataset_path: super_glue
+dataset_path: aps/super_glue
 dataset_name: wsc.fixed
 training_split: train
 validation_split: validation

lm_eval/tasks/triviaqa/README.md

@@ -12,7 +12,6 @@ high quality distant supervision for answering the questions.
 
 Homepage: https://nlp.cs.washington.edu/triviaqa/
 
-
 ### Citation
 
 ```
@@ -40,15 +39,18 @@ Homepage: https://nlp.cs.washington.edu/triviaqa/
 ### Checklist
 
 For adding novel benchmarks/datasets to the library:
-* [ ] Is the task an existing benchmark in the literature?
-  * [ ] Have you referenced the original paper that introduced the task?
-  * [ ] If yes, does the original paper provide a reference implementation? If so, have you checked against the reference implementation and documented how to run such a test?
 
+* [ ] Is the task an existing benchmark in the literature?
+    * [ ] Have you referenced the original paper that introduced the task?
+    * [ ] If yes, does the original paper provide a reference implementation? If so, have you checked against the
+      reference implementation and documented how to run such a test?
 
 If other tasks on this dataset are already supported:
+
 * [ ] Is the "Main" variant of this task clearly denoted?
 * [ ] Have you provided a short sentence in a README on what each new variant adds / evaluates?
 * [ ] Have you noted which, if any, published evaluation setups are matched by this variant?
 
 ### Changelog
-* 2025-07-21: Added `multi_target` to `exact_match`. Scores should not change.
+
+* 2025-07-21: Added `multiple_targets` to `exact_match`. Scores should not change.

lm_eval/tasks/triviaqa/default.yaml

@@ -6,6 +6,7 @@ training_split: train
 validation_split: validation
 doc_to_text: "Question: {{question}}?\nAnswer:"
 doc_to_target: "{{answer.aliases}}"
+multiple_targets: true
 should_decontaminate: true
 doc_to_decontamination_query: question
 generation_kwargs:
@@ -27,6 +28,6 @@ metric_list:
     higher_is_better: true
     ignore_case: true
     ignore_punctuation: true
-    multi_target: true
+    multiple_targets: true
 metadata:
   version: 3.0

lm_eval/tasks/winogrande/default.yaml

@@ -9,6 +9,7 @@ doc_to_target: !function preprocess_winogrande.doc_to_target
 doc_to_choice: !function preprocess_winogrande.doc_to_choice
 should_decontaminate: true
 doc_to_decontamination_query: sentence
+multiple_inputs: true
 metric_list:
   - metric: acc
     aggregation: mean

lm_eval/tasks/winogrande/preprocess_winogrande.py

-def doc_to_text(doc):
+def doc_to_target(doc) -> int:
     answer_to_num = {"1": 0, "2": 1}
     return answer_to_num[doc["answer"]]
 
 
-def doc_to_target(doc):
+def doc_to_choice(doc) -> list[str]:
     idx = doc["sentence"].index("_") + 1
-    return doc["sentence"][idx:].strip()
+    return [doc["sentence"][idx:].strip()]
 
 
-def doc_to_choice(doc):
+def doc_to_text(doc) -> list[str]:
     idx = doc["sentence"].index("_")
     options = [doc["option1"], doc["option2"]]
     return [doc["sentence"][:idx] + opt for opt in options]

lm_eval/utils.py

@@ -12,10 +12,9 @@ import os
 import re
 from collections.abc import Generator
 from dataclasses import asdict, is_dataclass
-from functools import lru_cache, partial, wraps
+from functools import wraps
 from itertools import islice
-from pathlib import Path
-from typing import Any, Callable, Dict, List, Optional
+from typing import Any, Callable
 
 import numpy as np
 from jinja2 import BaseLoader, Environment, StrictUndefined
@@ -27,7 +26,9 @@ HIGHER_IS_BETTER_SYMBOLS = {
     True: "↑",
     False: "↓",
 }
-def wrap_text(string: str, width: int = 140, **kwargs) -> Optional[str]:
+
+
+def wrap_text(string: str, width: int = 140, **kwargs) -> str | None:
     """
     Wraps the given string to the specified width.
     """
@@ -44,8 +45,7 @@ def wrap_text(string: str, width: int = 140, **kwargs) -> Optional[str]:
     )
 
 
-
-def get_logger(level: Optional[str] = None) -> logging.Logger:
+def get_logger(level: str | None = None) -> logging.Logger:
     """
     Get a logger with a stream handler that captures all lm_eval logs.
 
@@ -626,3 +626,7 @@ def apply_template(template: str, doc: dict) -> str:
         apply_template._env.filters["regex_replace"] = regex_replace
 
     return _compile_tpl(template).render(**doc)
+
+
+def validate_index(index: int, length: int) -> int:
+    return index if index < length else -100

tests/test_doc_to_methods.py

+"""Tests for ConfigurableTask doc_to_* methods with Jinja/YAML parsing.
+
+This test suite documents and validates all expected YAML input types for the doc_to_* methods:
+
+doc_to_text - Transforms a document into the input text for the model:
+  - String field name: References a field directly from the document
+    YAML: doc_to_text: "question"
+
+  - Jinja2 template: Renders a template with document fields
+    YAML: doc_to_text: "Question: {{question}}\nContext: {{context}}"
+
+  - Integer: Returns a constant integer value
+    YAML: doc_to_text: 0
+
+  - Python function: Applies a callable function (via !function directive)
+    YAML: doc_to_text: !function utils.my_custom_function
+
+doc_to_target - Transforms a document into the expected target/answer:
+  - String field name: References a field directly from the document
+    YAML: doc_to_target: "answer"
+
+  - Jinja2 template: Renders a template, can return string or int for multiple choice
+    YAML: doc_to_target: "{{answers[correct_idx]}}"
+    YAML: doc_to_target: "{{label}}" # "0", "1", etc. converted to int if doc_to_choice exists
+
+  - Integer: Returns a constant integer value (typically for multiple choice)
+    YAML: doc_to_target: 0
+
+  - List of templates: Returns multiple targets: list[str]
+    YAML: doc_to_target: ["{{answer1}}", "{{answer2}}"]
+
+  - Python function: Applies a callable function
+    YAML: doc_to_target: !function utils.extract_answer
+
+doc_to_choice - Defines the list of choices for multiple choice tasks:
+  - String field name: References a list field from the document
+    YAML: doc_to_choice: "options"
+
+  - Jinja2 template returning list: Template that evaluates to a list
+    YAML: doc_to_choice: "{{choices}}" # Must render to "['A', 'B', 'C']" format
+    YAML: doc_to_choice: "{{[correct, wrong]}}" # Creates list literal from fields
+    YAML: doc_to_choice: "{{options if options else default_options}}"
+
+  - List of templates: Each template becomes a choice
+    YAML: doc_to_choice: ["{{choice_a}}", "{{choice_b}}", "{{choice_c}}"]
+
+  - Dictionary: Values become the choices (keys are ignored)
+    YAML: doc_to_choice:
+      A: "First option"
+      B: "Second option"
+      C: "Third option"
+
+  - Python function: Returns a list of choices
+    YAML: doc_to_choice: !function utils.generate_choices
+
+Special Jinja2 features supported:
+  - Filters: {{text|upper}}, {{text|lower}}, {{text|regex_replace('pattern', 'replacement')}}
+  - Conditionals: {{field1 if condition else field2}}
+  - List operations: {{', '.join(items)}}
+  - Nested field access: {{metadata.answer}}, {{choices[0]}}
+  - Math operations: {{score * 100}}
+  - String concatenation: {{first + ' ' + last}}
+"""
+
+from unittest.mock import Mock, patch
+
+import pytest
+
+from lm_eval.api.task import ConfigurableTask
+
+
+class TestDocToTextMethod:
+    """Test suite for doc_to_text method."""
+
+    def test_doc_to_text_with_string_field(self):
+        """Test doc_to_text when config points to a field name."""
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.features = ["text", "answer", "choices", "label"]
+        task.config = Mock()
+        task.config.doc_to_text = "text"
+
+        doc = {"text": "This is a test question", "answer": "A"}
+
+        result = ConfigurableTask.doc_to_text(task, doc)
+        assert result == "This is a test question"
+
+    def test_doc_to_text_with_jinja_template(self):
+        """Test doc_to_text with Jinja template."""
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.features = ["text", "answer"]
+        task.config = Mock()
+        task.config.doc_to_text = "Question: {{text}}"
+
+        doc = {"text": "What is 2+2?", "answer": "4"}
+
+        result = ConfigurableTask.doc_to_text(task, doc)
+        assert result == "Question: What is 2+2?"
+
+    def test_doc_to_text_with_complex_jinja(self):
+        """Test doc_to_text with complex Jinja expressions."""
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.features = ["text", "answer"]
+        task.config = Mock()
+        task.config.doc_to_text = "{{text|upper}} - {{answer|lower}}"
+
+        doc = {"text": "Test", "answer": "ANSWER"}
+
+        result = ConfigurableTask.doc_to_text(task, doc)
+        assert result == "TEST - answer"
+
+    def test_doc_to_text_with_list(self):
+        """Test doc_to_text when config is an integer."""
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.config = Mock()
+        task.config.doc_to_text = ["{{choice1}}", "{{choice2}}"]
+
+        doc = {"choice1": "1", "choice2": "2"}
+
+        result = ConfigurableTask.doc_to_text(task, doc)
+        assert result == ["1", "2"]
+
+    def test_doc_to_text_with_callable(self):
+        """Test doc_to_text with a callable function."""
+
+        def custom_text_func(doc):
+            return f"Custom: {doc['text']}"
+
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.config = Mock()
+        task.config.doc_to_text = custom_text_func
+
+        doc = {"text": "test"}
+
+        result = ConfigurableTask.doc_to_text(task, doc)
+        assert result == "Custom: test"
+
+    def test_doc_to_text_with_regex_filter(self):
+        """Test doc_to_text with Jinja regex_replace filter."""
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.features = ["text"]
+        task.config = Mock()
+        task.config.doc_to_text = "{{text|regex_replace('\\d+', 'X')}}"
+
+        doc = {"text": "There are 123 apples and 456 oranges"}
+
+        result = ConfigurableTask.doc_to_text(task, doc)
+        assert result == "There are X apples and X oranges"
+
+    def test_doc_to_text_with_list_comprehension(self):
+        """Test doc_to_text with Jinja list comprehension."""
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_text = "Options: {{ ', '.join(choices) }}"
+
+        doc = {"choices": ["red", "green", "blue"]}
+
+        result = ConfigurableTask.doc_to_text(task, doc)
+        assert result == "Options: red, green, blue"
+
+    def test_override_doc_to_text(self):
+        """Test overriding doc_to_text with parameter."""
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_text = "default"
+
+        doc = {"text": "test"}
+
+        result = ConfigurableTask.doc_to_text(task, doc, doc_to_text="override")
+        assert result == "override"
+
+    def test_doc_to_text_type_error(self):
+        """Test doc_to_text raises TypeError for invalid type."""
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.config = Mock()
+        task.config.doc_to_text = {"invalid": "type"}
+
+        doc = {"text": "test"}
+
+        with pytest.raises(TypeError):
+            ConfigurableTask.doc_to_text(task, doc)
+
+    def test_doc_to_text_with_missing_field(self):
+        """Test doc_to_text with missing field in template."""
+        task = Mock(spec=ConfigurableTask)
+        task.multiple_inputs = False
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_text = "{{missing_field}}"
+
+        doc = {"text": "test"}
+
+        from jinja2 import UndefinedError
+
+        with pytest.raises(UndefinedError):
+            ConfigurableTask.doc_to_text(task, doc)
+
+
+class TestDocToTargetMethod:
+    """Test suite for doc_to_target method."""
+
+    def test_doc_to_target_with_field(self):
+        """Test doc_to_target when config points to a field name."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = ["text", "answer"]
+        task.config = Mock()
+        task.config.doc_to_target = "answer"
+        task._config = task.config
+
+        doc = {"text": "question", "answer": "correct answer"}
+
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == "correct answer"
+
+    def test_doc_to_target_with_jinja_template(self):
+        """Test doc_to_target with Jinja template."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_target = "{{answer}}"
+        task.config.doc_to_choice = None
+        task._config = task.config
+
+        doc = {"answer": "test_answer"}
+
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == "test_answer"
+
+    def test_doc_to_target_with_jinja_index(self):
+        """Test doc_to_target with Jinja template returning numeric string."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_target = "{{label}}"
+        task.config.doc_to_choice = ["A", "B", "C"]
+        task._config = task.config
+
+        doc = {"label": "1"}
+
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == 1  # Should be converted to int
+
+    def test_doc_to_target_with_int(self):
+        """Test doc_to_target when config is an integer."""
+        task = Mock(spec=ConfigurableTask)
+        task.config = Mock()
+        task.config.doc_to_target = 0
+        task._config = task.config
+
+        doc = {"answer": "test"}
+
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == 0
+
+    def test_doc_to_target_with_list(self):
+        """Test doc_to_target with list of templates."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_target = ["{{answer}}", "{{text}}"]
+        task._config = task.config
+
+        doc = {"answer": "A", "text": "question"}
+
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == ["A", "question"]
+
+    def test_doc_to_target_with_int_list(self):
+        """Test doc_to_target with list of templates."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.multiple_targets = True
+        task.config = Mock()
+        task.config.doc_to_target = "{{answer}}"
+        task._config = task.config
+
+        doc = {"answer": [1, 2, 3, 4]}
+
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == [1, 2, 3, 4]
+
+    def test_doc_to_target_with_callable(self):
+        """Test doc_to_target with a callable function."""
+
+        def custom_target_func(doc):
+            return doc["label"] * 2
+
+        task = Mock(spec=ConfigurableTask)
+        task.config = Mock()
+        task.config.doc_to_target = custom_target_func
+        task._config = task.config
+
+        doc = {"label": 3}
+
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == 6
+
+    def test_doc_to_target_with_nested_fields(self):
+        """Test doc_to_target with nested field access."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_target = "{{meta.answer}}"
+        task.config.doc_to_choice = None
+        task._config = task.config
+
+        doc = {"meta": {"answer": "nested_value"}}
+
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == "nested_value"
+
+    def test_doc_to_target_multiple_targets(self):
+        """Test doc_to_target returning list for multiple targets."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_target = ["{{answer1}}", "{{answer2}}"]
+        task._config = task.config
+
+        doc = {"answer1": "first", "answer2": "second"}
+
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == ["first", "second"]
+
+    def test_override_doc_to_target(self):
+        """Test overriding doc_to_target with parameter."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_target = "default"
+        task._config = task.config
+
+        doc = {"answer": "test"}
+
+        result = ConfigurableTask.doc_to_target(task, doc, doc_to_target="override")
+        assert result == "override"
+
+    def test_doc_to_target_type_error(self):
+        """Test doc_to_target raises TypeError for invalid type."""
+        task = Mock(spec=ConfigurableTask)
+        task.config = Mock()
+        task.config.doc_to_target = {"invalid": "type"}
+        task._config = task.config
+
+        doc = {"answer": "test"}
+
+        with pytest.raises(TypeError):
+            ConfigurableTask.doc_to_target(task, doc)
+
+    def test_doc_to_target_literal_eval_edge_cases(self):
+        """Test doc_to_target with edge cases for literal_eval."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_choice = ["A", "B", "C"]
+        task._config = task.config
+
+        # Test numeric string conversion
+        task.config.doc_to_target = "{{label}}"
+        doc = {"label": "2"}
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == 2
+
+        # Test non-numeric string stays as string
+        doc = {"label": "abc"}
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == "abc"
+
+        # Test mixed alphanumeric stays as string
+        doc = {"label": "2a"}
+        result = ConfigurableTask.doc_to_target(task, doc)
+        assert result == "2a"
+
+
+class TestDocToChoiceMethod:
+    """Test suite for doc_to_choice method."""
+
+    def test_doc_to_choice_with_field(self):
+        """Test doc_to_choice when config points to a field name."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = ["choices"]
+        task.config = Mock()
+        task.config.doc_to_choice = "choices"
+
+        doc = {"choices": ["A", "B", "C", "D"]}
+
+        result = ConfigurableTask.doc_to_choice(task, doc)
+        assert result == ["A", "B", "C", "D"]
+
+    def test_doc_to_choice_with_jinja_list(self):
+        """Test doc_to_choice with Jinja template returning list as string."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_choice = "{{choices}}"
+
+        doc = {"choices": ["opt1", "opt2", "opt3"]}
+
+        # The Jinja template will render the list as a string
+        result = ConfigurableTask.doc_to_choice(task, doc)
+        assert result == ["opt1", "opt2", "opt3"]
+
+    def test_doc_to_choice_with_jinja_list_literal(self):
+        """Test doc_to_choice with Jinja template creating a list literal."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_choice = "{{[correct, wrong]}}"
+
+        doc = {"correct": "The right answer", "wrong": "The wrong answer"}
+
+        # The Jinja template will create a list literal and render it as a string
+        result = ConfigurableTask.doc_to_choice(task, doc)
+        assert result == ["The right answer", "The wrong answer"]
+
+        # Test with another variation
+        task.config.doc_to_choice = "{{[option_a, option_b, option_c]}}"
+        doc = {"option_a": "Choice A", "option_b": "Choice B", "option_c": "Choice C"}
+        result = ConfigurableTask.doc_to_choice(task, doc)
+        assert result == ["Choice A", "Choice B", "Choice C"]
+
+    def test_doc_to_choice_with_list_of_templates(self):
+        """Test doc_to_choice with list of Jinja templates."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_choice = ["{{choice_a}}", "{{choice_b}}", "{{choice_c}}"]
+
+        doc = {"choice_a": "Apple", "choice_b": "Banana", "choice_c": "Cherry"}
+
+        result = ConfigurableTask.doc_to_choice(task, doc)
+        assert result == ["Apple", "Banana", "Cherry"]
+
+    def test_doc_to_choice_with_dict(self):
+        """Test doc_to_choice with dictionary config."""
+        task = Mock(spec=ConfigurableTask)
+        task.config = Mock()
+        task.config.doc_to_choice = {
+            "A": "First option",
+            "B": "Second option",
+            "C": "Third option",
+        }
+
+        doc = {}
+
+        result = ConfigurableTask.doc_to_choice(task, doc)
+        assert result == ["First option", "Second option", "Third option"]
+
+    def test_doc_to_choice_with_callable(self):
+        """Test doc_to_choice with a callable function."""
+
+        def custom_choice_func(doc):
+            return [f"Option {i}" for i in range(doc["num_choices"])]
+
+        task = Mock(spec=ConfigurableTask)
+        task.config = Mock()
+        task.config.doc_to_choice = custom_choice_func
+
+        doc = {"num_choices": 3}
+
+        result = ConfigurableTask.doc_to_choice(task, doc)
+        assert result == ["Option 0", "Option 1", "Option 2"]
+
+    def test_doc_to_choice_none_error(self):
+        """Test doc_to_choice logs error when not configured."""
+        task = Mock(spec=ConfigurableTask)
+        task.config = Mock()
+        task.config.doc_to_choice = None
+
+        doc = {}
+
+        # When doc_to_choice is None, it logs an error and then raises TypeError
+        with patch("lm_eval.api.task.eval_logger.error") as mock_error:
+            with pytest.raises(TypeError):
+                ConfigurableTask.doc_to_choice(task, doc)
+            mock_error.assert_called_once_with(
+                "doc_to_choice was called but not set in config"
+            )
+
+    def test_doc_to_choice_with_conditional(self):
+        """Test doc_to_choice with Jinja conditional."""
+        task = Mock(spec=ConfigurableTask)
+        task.features = []
+        task.config = Mock()
+        task.config.doc_to_choice = "{{choices if has_choices else default_choices}}"
+
+        doc = {
+            "has_choices": True,
+            "choices": ["A", "B"],
+            "default_choices": ["X", "Y"],
+        }
+
+        result = ConfigurableTask.doc_to_choice(task, doc)
+        assert result == ["A", "B"]
+
+    def test_override_doc_to_choice(self):
+        """Test overriding doc_to_choice with parameter."""
+        task = Mock(spec=ConfigurableTask)
+        task.config = Mock()
+        task.config.doc_to_choice = ["A", "B"]
+
+        doc = {}
+
+        result = ConfigurableTask.doc_to_choice(
+            task, doc, doc_to_choice=["X", "Y", "Z"]
+        )
+        assert result == ["X", "Y", "Z"]
+
+    def test_doc_to_choice_type_error(self):
+        """Test doc_to_choice raises TypeError for invalid type."""
+        task = Mock(spec=ConfigurableTask)
+        task.config = Mock()
+        task.config.doc_to_choice = 123  # Invalid type
+
+        doc = {}
+
+        with pytest.raises(TypeError):
+            ConfigurableTask.doc_to_choice(task, doc)

tests/test_tasks.py

@@ -29,7 +29,7 @@ def get_new_tasks_else_default():
     return task_classes if task_classes else TASKS
 
 
-def task_class(task_names=None, task_manager=None) -> ConfigurableTask:
+def task_class(task_names=None, task_manager=None) -> list[ConfigurableTask]:
     """
     Convert a list of task names to a list of ConfigurableTask instances
     """
@@ -98,7 +98,7 @@ class TestBaseTasks:
         _array = [task.doc_to_text(doc) for doc in arr]
         # space convention; allow txt to have length 0 for perplexity-like tasks since the model tacks an <|endoftext|> on
         target_delimiter: str = task.config.target_delimiter
-        if not task.multiple_input:
+        if not task.multiple_inputs:
             for x in _array:
                 assert isinstance(x, str)
                 assert (
@@ -133,10 +133,7 @@ class TestBaseTasks:
         _array_target = [task.doc_to_target(doc) for doc in arr]
         if task._config.output_type == "multiple_choice":
             # TODO<baber>: label can be string or int; add better test conditions
-            assert all(
-                (isinstance(label, int) or isinstance(label, str))
-                for label in _array_target
-            )
+            assert all(isinstance(label, (int, str)) for label in _array_target)
 
     def test_build_all_requests(self, task_class, limit):
         task_class.build_all_requests(rank=1, limit=limit, world_size=1)
@@ -153,7 +150,7 @@ class TestBaseTasks:
         # ctx is "" for multiple input tasks
         requests = [
             task.construct_requests(
-                doc=doc, ctx="" if task.multiple_input else task.doc_to_text(doc)
+                doc=doc, ctx=[""] if task.multiple_inputs else task.doc_to_text(doc)
             )
             for doc in arr
         ]
@@ -187,7 +184,7 @@ class TestUnitxtTasks(TestBaseTasks):
     """
 
     def test_check_training_docs(self, task_class: ConfigurableTask):
-        if task_class.has_training_docs():
+        if task_class.has_training_docs:
             assert task_class.dataset["train"] is not None
 
     def test_check_validation_docs(self, task_class):