include ranges rather than sampled values in difficulty metadata dicts (#387)

* update difficulty metadata for logic datasets

* update difficulty metadata for graph datasets

* update difficulty metadata for geometry datasets

* update difficulty metadata for games datasets

* update difficulty metadata for cognition datasets

* update difficulty metadata for arithmetic datasets

* update difficulty metadata for arc datasets

* update difficulty metadata for algorithmic datasets

* update difficulty metadata for algebra datasets

* use tuples

* update tests

* update tests

tests/test_boxnet.py

@@ -43,8 +43,8 @@ def test_boxnet_items():
         assert "initial_state" in item["metadata"]
 
         # Verify row_num and column_num are within limits
-        row_num = item["metadata"]["difficulty"]["row_num"]
-        column_num = item["metadata"]["difficulty"]["column_num"]
+        row_num = item["metadata"]["row_num"]
+        column_num = item["metadata"]["column_num"]
         assert 1 <= row_num <= 2, f"row_num {row_num} outside valid range"
         assert 1 <= column_num <= 2, f"column_num {column_num} outside valid range"
 
@@ -78,8 +78,8 @@ def test_boxnet_grid_sizes():
 
     for i in range(len(dataset)):
         item = dataset[i]
-        row_num = item["metadata"]["difficulty"]["row_num"]
-        column_num = item["metadata"]["difficulty"]["column_num"]
+        row_num = item["metadata"]["row_num"]
+        column_num = item["metadata"]["column_num"]
 
         rows_set.add(row_num)
         columns_set.add(column_num)

tests/test_coaching.py

@@ -53,11 +53,15 @@ def test_coach_with_chain_sum():
     # Each key should be a tuple of tuples containing difficulty parameters
     for key in aggregated.scores:
         assert isinstance(key, tuple)
-        # Each inner tuple should be (param_name, value)
+        # Each inner tuple should be (param_name, value) or (param_name, (min_value, max_value))
         for param in key:
             assert isinstance(param, tuple)
             assert param[0] in ("num_terms", "num_digits")
-            assert isinstance(param[1], int)
+            assert (
+                isinstance(param[1], int)
+                or (isinstance(param[1], tuple) and len(param[1]) == 2)
+                and all(isinstance(v, int) for v in param[1])
+            )
 
     # Test aggregation with last_n
     last_3 = coach.score_board.aggregate(last_n=3)
@@ -171,7 +175,7 @@ def test_coach_with_composite():
         item = coach[i + 5]  # Use different indices
         if "chain_sum" in item["metadata"]["source_dataset"]:
             metadata = item["metadata"]
-            assert metadata["difficulty"]["num_terms"] >= 4
+            assert metadata["num_terms"] >= 4
 
 
 def test_grouped_scores_str():

tests/test_rearc.py

@@ -38,12 +38,12 @@ def test_rearc_items():
         assert "input" in meta
         assert "output" in meta
         assert "task_id" in meta
-        assert "rng" in meta["difficulty"]
-        assert "pso" in meta["difficulty"]
+        assert "rng" in meta
+        assert "pso" in meta
 
         # Validate difficulty bounds
-        assert config.diff_lb <= meta["difficulty"]["rng"] <= config.diff_ub
-        assert config.diff_lb <= meta["difficulty"]["pso"] <= config.diff_ub
+        assert config.diff_lb <= meta["rng"] <= config.diff_ub
+        assert config.diff_lb <= meta["pso"] <= config.diff_ub
 
 
 def test_rearc_solution_validation():

tests/test_tsumego.py

@@ -124,7 +124,7 @@ def test_score_answer():
 
     # test optimal score for answers, patching each entry
     for x in dataset:
-        assert len(x["metadata"]["board"]) == x["metadata"]["difficulty"]["board_size"]
+        assert len(x["metadata"]["board"]) == x["metadata"]["board_size"]
         assert dataset.score_answer(x["answer"], entry=x) == 1.0