Generate Japanese and Ukrainian splits, update configs and readme

README.md

@@ -9,9 +9,11 @@ language:
 - is
 - nb
 - it
+- ja
 - mr
 - nl
 - ru
+- uk
 - zh
 license: mit
 pretty_name: Multilingual GSM-Symbolic
@@ -105,6 +107,18 @@ configs:
     path: data/zho/test_original-*.parquet
   - split: test_synthetic
     path: data/zho/test_synthetic-*.parquet
+- config_name: jpn
+  data_files:
+  - split: test_original
+    path: data/jpn/test_original-*.parquet
+  - split: test_synthetic
+    path: data/jpn/test_synthetic-*.parquet
+- config_name: ukr
+  data_files:
+  - split: test_original
+    path: data/ukr/test_original-*.parquet
+  - split: test_synthetic
+    path: data/ukr/test_synthetic-*.parquet
 - config_name: eng_metric
   data_files:
   - split: test_original

data/generation_log.json

@@ -133,5 +133,29 @@
     ],
     "test_n_original": 100,
     "test_n_synthetic": 2000
+  },
+  "jpn": {
+    "seed": 42,
+    "n_synthetic_per_template": 20,
+    "commit": "a5b3b94fb239d8c74f75dabe5780d23dccc63dfb",
+    "package_version": "0.4.13",
+    "generated_at": "2026-07-08T10:18:33.734141+00:00",
+    "splits": [
+      "test"
+    ],
+    "test_n_original": 100,
+    "test_n_synthetic": 2000
+  },
+  "ukr": {
+    "seed": 42,
+    "n_synthetic_per_template": 20,
+    "commit": "a5b3b94fb239d8c74f75dabe5780d23dccc63dfb",
+    "package_version": "0.4.13",
+    "generated_at": "2026-07-08T10:25:29.664636+00:00",
+    "splits": [
+      "test"
+    ],
+    "test_n_original": 100,
+    "test_n_synthetic": 2000
   }
 }

data/jpn/test_original-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0a7be3a34a5a41f6d5579eebc92d6d88487de770c1f54b4f87a333df970dbfb
+size 42695

data/jpn/test_synthetic-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c7f8fa3faf523353fcd2484f0f02e2e705318cc7018686ac0d1a49a39725c97
+size 330579

data/ukr/test_original-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb38a444c12230f9890b5864a13adc4809a033d9d473cc9145801fc97a64f0a9
+size 50950

data/ukr/test_synthetic-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b3806d651c066ef8fbc0e0463a9d06ba85907eab76a7e5ccbd76bba9677ad6f
+size 392881

eval.yaml

@@ -593,3 +593,87 @@ tasks:
 
     scorers:
       - name: math
+
+  - id: original_jpn
+    config: jpn
+    split: test_original
+
+    field_spec:
+      input: question
+      target: target
+      metadata:
+        - answer
+        - language
+        - source_id
+
+    solvers:
+      - name: prompt_template
+        args:
+          template: "次の数学の問題をステップ・バイ・ステップで解いてください。最終的な数値の回答（単位、通貨記号、パーセント記号なし）を \\boxed{{}} の中に入れてください。例えば、\\boxed{{42}} のようにしてください。\n\n{prompt}"
+      - name: generate
+
+    scorers:
+      - name: math
+
+  - id: synthetic_jpn
+    config: jpn
+    split: test_synthetic
+
+    field_spec:
+      input: question
+      target: target
+      metadata:
+        - answer
+        - language
+        - source_id
+
+    solvers:
+      - name: prompt_template
+        args:
+          template: "次の数学の問題をステップ・バイ・ステップで解いてください。最終的な数値の回答（単位、通貨記号、パーセント記号なし）を \\boxed{{}} の中に入れてください。例えば、\\boxed{{42}} のようにしてください。\n\n{prompt}"
+      - name: generate
+
+    scorers:
+      - name: math
+
+  - id: original_ukr
+    config: ukr
+    split: test_original
+
+    field_spec:
+      input: question
+      target: target
+      metadata:
+        - answer
+        - language
+        - source_id
+
+    solvers:
+      - name: prompt_template
+        args:
+          template: "Розв'яжіть наступну математичну задачу крок за кроком. Запишіть остаточну числову відповідь (без одиниць виміру, символів валюти чи знаків відсотка) у \\boxed{{}}, наприклад: \\boxed{{42}}.\n\n{prompt}"
+      - name: generate
+
+    scorers:
+      - name: math
+
+  - id: synthetic_ukr
+    config: ukr
+    split: test_synthetic
+
+    field_spec:
+      input: question
+      target: target
+      metadata:
+        - answer
+        - language
+        - source_id
+
+    solvers:
+      - name: prompt_template
+        args:
+          template: "Розв'яжіть наступну математичну задачу крок за кроком. Запишіть остаточну числову відповідь (без одиниць виміру, символів валюти чи знаків відсотка) у \\boxed{{}}, наприклад: \\boxed{{42}}.\n\n{prompt}"
+      - name: generate
+
+    scorers:
+      - name: math

generate_hf_dataset.py

@@ -68,6 +68,9 @@ def _extract_target(answer: str) -> str:
 
 
 def build_original_rows(lang: str, split: str = "test") -> list[dict]:
+    split_dir = TEMPLATE_DATA_ROOT / lang / split
+    if split == "test" and not split_dir.exists():
+        split_dir = TEMPLATE_DATA_ROOT / lang / "symbolic"
     return [
         {
             "question": template.question,
@@ -76,13 +79,16 @@ def build_original_rows(lang: str, split: str = "test") -> list[dict]:
             "language": lang,
             "source_id": template.id_orig,
         }
-        for template in load_data(lang, directory=TEMPLATE_DATA_ROOT / lang / split)
+        for template in load_data(lang, directory=split_dir)
     ]
 
 
 def build_synthetic_rows(lang: str, n: int, rng: Random, split: str = "test") -> list[dict]:
     replacements = load_replacements(lang)
-    templates = load_data(lang, directory=TEMPLATE_DATA_ROOT / lang / split)
+    split_dir = TEMPLATE_DATA_ROOT / lang / split
+    if split == "test" and not split_dir.exists():
+        split_dir = TEMPLATE_DATA_ROOT / lang / "symbolic"
+    templates = load_data(lang, directory=split_dir)
     rows = []
     for i, template in enumerate(templates, 1):
         questions = template.generate_questions(n=n, replacements=replacements, rng=rng)
@@ -106,7 +112,11 @@ def generate(lang: str, seed: int, n: int, splits: list[str]) -> None:
 
     dataset = DatasetDict()
     split_counts = {}
-    available = [candidate for candidate in VALID_DATA_SPLITS if (TEMPLATE_DATA_ROOT / lang / candidate).exists()]
+    available = []
+    if (TEMPLATE_DATA_ROOT / lang / "test").exists() or (TEMPLATE_DATA_ROOT / lang / "symbolic").exists():
+        available.append("test")
+    if (TEMPLATE_DATA_ROOT / lang / "train").exists():
+        available.append("train")
     for split in splits:
         if split not in available:
             print(f"[{lang}] Split '{split}' is not available for this language.")