letter counting curriculum (#312)

reasoning_gym/algorithmic/__init__.py

@@ -19,7 +19,7 @@ from .graph_color import GraphColorConfig, GraphColorCurriculum, GraphColorDatas
 from .group_anagrams import GroupAnagramsConfig, GroupAnagramsCurriculum, GroupAnagramsDataset
 from .isomorphic_strings import IsomorphicStringsConfig, IsomorphicStringsCurriculum, IsomorphicStringsDataset
 from .jugs import JugsConfig, JugsDataset
-from .letter_counting import LetterCountingConfig, LetterCountingDataset
+from .letter_counting import LetterCountingConfig, LetterCountingCurriculum, LetterCountingDataset
 from .letter_jumble import LetterJumbleConfig, LetterJumbleDataset
 from .manipulate_matrix import ManipulateMatrixConfig, ManipulateMatrixCurriculum, ManipulateMatrixDataset
 from .number_filtering import NumberFilteringConfig, NumberFilteringDataset
@@ -66,6 +66,7 @@ __all__ = [
     "GameOfLifeHaltingDataset",
     "LetterCountingConfig",
     "LetterCountingDataset",
+    "LetterCountingCurriculum",
     "LetterJumbleConfig",
     "LetterJumbleDataset",
     "NumberFilteringConfig",

reasoning_gym/algorithmic/letter_counting.py

@@ -7,6 +7,7 @@ from typing import Optional
 
 from reasoning_gym.data import read_data_file
 
+from ..coaching import AttributeType, BaseCurriculum, RangeAttributeDefinition
 from ..factory import ProceduralDataset, register_dataset
 
 
@@ -41,7 +42,10 @@ class LetterCountingDataset(ProceduralDataset):
         rng = Random(self.seed + idx)
 
         # Select random span of words
-        span_length = rng.randint(self.config.min_words, self.config.max_words)
+        span_length = min(
+            rng.randint(self.config.min_words, self.config.max_words),
+            len(self.words),
+        )
         start_idx = rng.randint(0, len(self.words) - span_length)
         span = self.words[start_idx : start_idx + span_length]
 
@@ -59,8 +63,32 @@ class LetterCountingDataset(ProceduralDataset):
         return {
             "question": f'How many times does the letter "{target_letter}" appear in the text: "{" ".join(span)}"?',
             "answer": str(count),
-            "metadata": {"span_length": span_length, "target_letter": target_letter, "span": span},
+            "metadata": {
+                "span_length": span_length,
+                "target_letter": target_letter,
+                "span": span,
+                "difficulty": {"words": span_length},
+            },
         }
 
 
-register_dataset("letter_counting", LetterCountingDataset, LetterCountingConfig)
+class LetterCountingCurriculum(BaseCurriculum):
+    def __init__(self):
+        super().__init__(LetterCountingCurriculum.__name__, LetterCountingConfig)
+
+        # Define attributes
+        self._define_attributes(
+            RangeAttributeDefinition(
+                name="words",
+                levels=[10, 50, 100, 1000],
+                default_level=1,
+                description="Number of words in the span",
+                attr_type=AttributeType.APPEND,
+                min_value=1,
+                lower_field_name="min_words",
+                upper_field_name="max_words",
+            ),
+        )
+
+
+register_dataset("letter_counting", LetterCountingDataset, LetterCountingConfig, LetterCountingCurriculum)

tests/test_letter_counting.py

@@ -2,7 +2,11 @@
 
 import pytest
 
-from reasoning_gym.algorithmic.letter_counting import LetterCountingConfig, LetterCountingDataset
+from reasoning_gym.algorithmic.letter_counting import (
+    LetterCountingConfig,
+    LetterCountingCurriculum,
+    LetterCountingDataset,
+)
 
 
 def test_letter_counting_config_validation():
@@ -76,3 +80,24 @@ def test_letter_counting_text_preprocessing():
     assert len(dataset.words) > 0
     # Verify words contain only word characters
     assert all(word.isalnum() for word in dataset.words)
+
+
+def test_letter_counting_curriculum():
+    curriculum = LetterCountingCurriculum()
+
+    base_value = {"size": 150, "seed": 1}
+
+    base_cfg: LetterCountingConfig = curriculum.generate_configuration(base_value)
+    assert base_cfg.seed == 1
+    assert base_cfg.size == 150
+    assert base_cfg.min_words == 10 and base_cfg.max_words == 50
+
+    # test incrementing attribute levels
+    curriculum.increment_attr_level("words")
+    increased_cfg = curriculum.generate_configuration(base_value)
+    assert increased_cfg.min_words == 10 and increased_cfg.max_words == 100
+
+    # test decrementing attribute level for words again
+    curriculum.decrement_attr_level("words")
+    partially_decreased_cfg = curriculum.generate_configuration(base_value)
+    assert partially_decreased_cfg.min_words == 10 and partially_decreased_cfg.max_words == 50