diff --git a/reasoning_gym/games/__init__.py b/reasoning_gym/games/__init__.py
index a17d0ea6..2b01898e 100644
--- a/reasoning_gym/games/__init__.py
+++ b/reasoning_gym/games/__init__.py
@@ -7,7 +7,7 @@ Game tasks for training reasoning capabilities:
 """
 
 from .countdown import CountdownConfig, CountdownDataset
-from .emoji_mystery import EmojiMysteryConfig, EmojiMysteryDataset
+from .emoji_mystery import EmojiMysteryConfig, EmojiMysteryCurriculum, EmojiMysteryDataset
 from .futoshiki import FutoshikiConfig, FutoshikiDataset
 from .knight_swap import KnightSwapConfig, KnightSwapDataset
 from .mahjong import MahjongPuzzleConfig, MahjongPuzzleCurriculum, MahjongPuzzleDataset
@@ -25,6 +25,7 @@ __all__ = [
     "CountdownConfig",
     "CountdownDataset",
     "EmojiMysteryConfig",
+    "EmojiMysteryCurriculum",
     "EmojiMysteryDataset",
     "FutoshikiConfig",
     "FutoshikiDataset",
diff --git a/reasoning_gym/games/emoji_mystery.py b/reasoning_gym/games/emoji_mystery.py
index 27f48d87..be34b140 100644
--- a/reasoning_gym/games/emoji_mystery.py
+++ b/reasoning_gym/games/emoji_mystery.py
@@ -3,6 +3,7 @@ from dataclasses import dataclass
 from random import Random
 from typing import Any, Optional
 
+from ..coaching import AttributeType, BaseCurriculum, RangeAttributeDefinition
 from ..data import read_data_file
 from ..factory import ProceduralDataset, register_dataset
 
@@ -188,7 +189,14 @@ class EmojiMysteryDataset(ProceduralDataset):
         secret_sentence = rng.choice(self.sentences).strip().replace("\n", " ")
         encoded_sentence = self.encode(secret_sentence, secret_emoji)
         question = QUESTION_TEMPLATE.format(sentence=encoded_sentence, hint_function=hint_function)
-        return {"question": question, "answer": secret_sentence, "metadata": {"emoji": secret_emoji}}
+        return {
+            "question": question,
+            "answer": secret_sentence,
+            "metadata": {
+                "emoji": secret_emoji,
+                "difficulty": {"num_words_in_sentence": len(re.findall(r"\b\w+\b", secret_sentence))},
+            },
+        }
 
     def variance_selector_to_byte(self, variation_selector: str) -> Optional[int]:
         variation_selector_codepoint = ord(variation_selector)
@@ -233,4 +241,22 @@ class EmojiMysteryDataset(ProceduralDataset):
         return reward
 
 
-register_dataset("emoji_mystery", EmojiMysteryDataset, EmojiMysteryConfig)
+class EmojiMysteryCurriculum(BaseCurriculum):
+    def __init__(self):
+        super().__init__(EmojiMysteryCurriculum.__name__, EmojiMysteryConfig)
+
+        self._define_attributes(
+            RangeAttributeDefinition(
+                name="num_words_in_sentence",
+                levels=[3, 10, 20, 35],
+                default_level=0,
+                description="Number of words in the sentence",
+                attr_type=AttributeType.STATIC,
+                min_value=3,
+                lower_field_name="min_words_in_sentence",
+                upper_field_name="max_words_in_sentence",
+            ),
+        )
+
+
+register_dataset("emoji_mystery", EmojiMysteryDataset, EmojiMysteryConfig, EmojiMysteryCurriculum)
diff --git a/tests/test_emoji_mystery.py b/tests/test_emoji_mystery.py
index c7618240..4d754f45 100644
--- a/tests/test_emoji_mystery.py
+++ b/tests/test_emoji_mystery.py
@@ -2,7 +2,7 @@ from random import Random
 
 import pytest
 
-from reasoning_gym.games.emoji_mystery import EmojiMysteryConfig, EmojiMysteryDataset
+from reasoning_gym.games.emoji_mystery import EmojiMysteryConfig, EmojiMysteryCurriculum, EmojiMysteryDataset
 
 
 def test_emoji_mystery_config_validation():
@@ -101,3 +101,35 @@ def test_emoji_mystery_scoring():
 
     # Test None answer
     assert dataset.score_answer(None, entry) == 0.0
+
+
+def test_emoji_mystery_curriculum():
+    """Test the emoji mystery curriculum functionality"""
+    curriculum = EmojiMysteryCurriculum()
+
+    base_value = {"size": 150, "seed": 1}
+
+    # Test base configuration
+    base_cfg: EmojiMysteryConfig = curriculum.generate_configuration(base_value)
+    assert base_cfg.seed == 1
+    assert base_cfg.size == 150
+    assert base_cfg.min_words_in_sentence == 3
+    assert base_cfg.max_words_in_sentence == 3
+
+    # Test incrementing attribute level
+    curriculum.increment_attr_level("num_words_in_sentence")
+    increased_cfg = curriculum.generate_configuration(base_value)
+    assert increased_cfg.min_words_in_sentence == 10
+    assert increased_cfg.max_words_in_sentence == 10
+
+    # Test incrementing attribute level again
+    curriculum.increment_attr_level("num_words_in_sentence")
+    double_increased_cfg = curriculum.generate_configuration(base_value)
+    assert double_increased_cfg.min_words_in_sentence == 20
+    assert double_increased_cfg.max_words_in_sentence == 20
+
+    # Test decrementing attribute level
+    curriculum.decrement_attr_level("num_words_in_sentence")
+    decreased_cfg = curriculum.generate_configuration(base_value)
+    assert decreased_cfg.min_words_in_sentence == 10
+    assert decreased_cfg.max_words_in_sentence == 10