diff --git a/reasoning_gym/algorithmic/sentence_reordering.py b/reasoning_gym/algorithmic/sentence_reordering.py index 069dda7c..57f19d6e 100644 --- a/reasoning_gym/algorithmic/sentence_reordering.py +++ b/reasoning_gym/algorithmic/sentence_reordering.py @@ -102,23 +102,16 @@ class SentenceReorderingDataset(ProceduralDataset): goal_words = expected_answer.split() answer_words = answer.split() if len(goal_words) == len(answer_words): - credit = [1 if goal_word.lower() == answer_word.lower() else 0 for goal_word, answer_word in zip(goal_words, answer_words)] + credit = [ + 1 if goal_word.lower() == answer_word.lower() else 0 + for goal_word, answer_word in zip(goal_words, answer_words) + ] reward = sum(credit) / len(credit) else: reward = 0.05 except: reward = 0.01 return reward - - - - - - - - - - register_dataset("sentence_reordering", SentenceReorderingDataset, SentenceReorderingConfig) diff --git a/reasoning_gym/algorithmic/word_ladder.py b/reasoning_gym/algorithmic/word_ladder.py index 40e36291..3be99138 100644 --- a/reasoning_gym/algorithmic/word_ladder.py +++ b/reasoning_gym/algorithmic/word_ladder.py @@ -8,7 +8,6 @@ from typing import Dict, List, Optional, Set, Tuple from ..data import get_data_file_path from ..factory import ProceduralDataset, register_dataset - QUESTION_TEMPLATE = """Transform the word ladder '{start}' to '{end}' by changing one letter at a time. Provide your answer as a comma-separated sequence of uppercase letters without spaces. Each step must be a valid English word.""" diff --git a/reasoning_gym/arithmetic/basic_arithmetic.py b/reasoning_gym/arithmetic/basic_arithmetic.py index efe9d465..c72ff143 100644 --- a/reasoning_gym/arithmetic/basic_arithmetic.py +++ b/reasoning_gym/arithmetic/basic_arithmetic.py @@ -225,16 +225,11 @@ class BasicArithmeticDataset(ProceduralDataset): def _format_question(self, rng: Random, expression: str) -> str: """Format the expression with clear answer positioning""" answer_instruction = "Put your final answer after '=' without additional text." - + if self.config.format_style == "simple": return f"{answer_instruction} Calculate {expression} =" else: - templates = [ - "What is {0} =", - "Solve {0}=", - "Compute {0} =", - "Evaluate: {0} =" - ] + templates = ["What is {0} =", "Solve {0}=", "Compute {0} =", "Evaluate: {0} ="] template = rng.choice(templates).format(expression) return f"{answer_instruction} {template}" diff --git a/tests/test_sentence_reordering.py b/tests/test_sentence_reordering.py index 05645c70..9348ec04 100644 --- a/tests/test_sentence_reordering.py +++ b/tests/test_sentence_reordering.py @@ -37,7 +37,7 @@ def test_getitem(dataset, config): assert "metadata" in item assert item["metadata"]["word_count"] >= config.min_words_in_sentence assert item["metadata"]["word_count"] <= config.max_words_in_sentence - assert len(item['answer'].split()) == item['metadata']['word_count'] + assert len(item["answer"].split()) == item["metadata"]["word_count"] def test_key_error_in_getitem(dataset):