Normalize eukaryote_cls schema and metadata

README.md

@@ -56,6 +56,28 @@ dataset_info:
     num_examples: 2600
   download_size: 4880581
   dataset_size: 6130755
+- config_name: eukaryote_cls
+  features:
+  - name: id
+    dtype: int64
+  - name: subtask
+    dtype: string
+  - name: question
+    dtype: string
+  - name: options
+    list: string
+  - name: answer_index
+    dtype: int64
+  - name: answer_text
+    dtype: string
+  - name: metadata
+    dtype: string
+  splits:
+  - name: train
+    num_bytes: 21610676
+    num_examples: 1000
+  download_size: 16997912
+  dataset_size: 21610676
 - config_name: binary_600
   features:
   - name: id
@@ -284,6 +306,8 @@ headline score is needed.
 | `mcq_prokaryote` | 165 | MCQ `3_species_cls` questions whose target organism is prokaryotic. |
 | `eukaryote_cls` | 1,000 | Auxiliary four-option eukaryotic taxonomy set stored separately from the original 5,200-example benchmark. |
 
+The `eukaryote_cls` configuration uses the `mcq` feature schema, including JSON-serialized `metadata`.
+
 The `binary_600` and `mcq_600` configurations were sampled without replacement using seed 42 from source revision
 `a71ab7e0e72b8eca7acd4475744dccd0159b20be`. The 100 selected examples from each subtask were then combined and
 shuffled so the rows are not stored in subtask blocks. These subsets are balanced by task family, but not necessarily
@@ -330,3 +354,14 @@ eukaryote_mcq = load_dataset("HuggingFaceBio/genomeqa", "mcq_eukaryote", split="
   year={2026}
 }
 ```
+
+## Reproducing
+
+To rebuild and publish the normalized `eukaryote_cls` configuration from the pinned source revision, run:
+
+```bash
+python scripts/create_eukaryote_cls.py --source-revision 18f68fabbcd43e8c227869a833027c969b39d7fe --output-dataset HuggingFaceBio/genomeqa --publish
+```
+
+The script preserves the source rows, converts `metadata` from Python dictionary strings to JSON, and casts the
+columns to the `mcq` feature schema. It publishes the Parquet file, this card, and the builder script in one commit.

mcq/eukaryote_cls_mcq.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f7f97e1f9e83b0a1af9bfe3543fd5ae8cabd0fb8147931b1aa7a8a95ae53c91
-size 9839351
+oid sha256:594ad54b23e4998896a11b99da23e8392cdeb483c6af648d0389b8c2f96d23e7
+size 16997912

scripts/create_eukaryote_cls.py

@@ -0,0 +1,99 @@
+"""Normalize the GenomeQA eukaryote_cls configuration to the mcq schema."""
+
+import argparse
+import ast
+import json
+from pathlib import Path
+
+from datasets import load_dataset
+from huggingface_hub import CommitOperationAdd, HfApi, hf_hub_download
+
+
+DATASET_ID = "HuggingFaceBio/genomeqa"
+PARQUET_PATH = "mcq/eukaryote_cls_mcq.parquet"
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--source-revision", required=True)
+    parser.add_argument("--output-dataset", default=DATASET_ID)
+    parser.add_argument("--output-dir", type=Path, default=Path.cwd())
+    parser.add_argument("--publish", action="store_true")
+    args = parser.parse_args()
+
+    reference = load_dataset(DATASET_ID, "mcq", split="train", revision=args.source_revision)
+    source = load_dataset(DATASET_ID, "eukaryote_cls", split="train", revision=args.source_revision)
+    normalized = source.map(lambda row: {"metadata": json.dumps(ast.literal_eval(row["metadata"]))})
+    normalized = normalized.cast(reference.features)
+
+    parquet_path = args.output_dir / PARQUET_PATH
+    parquet_path.parent.mkdir(parents=True, exist_ok=True)
+    normalized.to_parquet(parquet_path)
+
+    api = HfApi()
+    head = api.repo_info(args.output_dataset, repo_type="dataset").sha
+    card = Path(hf_hub_download(args.output_dataset, "README.md", repo_type="dataset", revision=head)).read_text()
+    if "- config_name: eukaryote_cls\n  features:" not in card:
+        info = f"""- config_name: eukaryote_cls
+  features:
+  - name: id
+    dtype: int64
+  - name: subtask
+    dtype: string
+  - name: question
+    dtype: string
+  - name: options
+    list: string
+  - name: answer_index
+    dtype: int64
+  - name: answer_text
+    dtype: string
+  - name: metadata
+    dtype: string
+  splits:
+  - name: train
+    num_bytes: {normalized.data.nbytes}
+    num_examples: {len(normalized)}
+  download_size: {parquet_path.stat().st_size}
+  dataset_size: {normalized.data.nbytes}
+"""
+        card = card.replace("- config_name: binary_600\n", info + "- config_name: binary_600\n", 1)
+
+    if "The `eukaryote_cls` configuration uses the `mcq` feature schema" not in card:
+        note = "The `eukaryote_cls` configuration uses the `mcq` feature schema, including JSON-serialized `metadata`.\n\n"
+        card = card.replace("The `binary_600` and `mcq_600` configurations", note + "The `binary_600` and `mcq_600` configurations", 1)
+
+    if "scripts/create_eukaryote_cls.py" not in card:
+        card += f"""
+## Reproducing
+
+To rebuild and publish the normalized `eukaryote_cls` configuration from the pinned source revision, run:
+
+```bash
+python scripts/create_eukaryote_cls.py --source-revision {args.source_revision} --output-dataset {args.output_dataset} --publish
+```
+
+The script preserves the source rows, converts `metadata` from Python dictionary strings to JSON, and casts the
+columns to the `mcq` feature schema. It publishes the Parquet file, this card, and the builder script in one commit.
+"""
+
+    card_path = args.output_dir / "README.md"
+    card_path.write_text(card)
+    print(f"Prepared {len(normalized)} rows at {parquet_path}")
+    if args.publish:
+        commit = api.create_commit(
+            repo_id=args.output_dataset,
+            repo_type="dataset",
+            parent_commit=head,
+            commit_message="Normalize eukaryote_cls schema and metadata",
+            operations=[
+                CommitOperationAdd(path_in_repo=PARQUET_PATH, path_or_fileobj=parquet_path),
+                CommitOperationAdd(path_in_repo="README.md", path_or_fileobj=card_path),
+                CommitOperationAdd(path_in_repo="scripts/create_eukaryote_cls.py", path_or_fileobj=__file__),
+            ],
+        )
+        print(f"Published {commit.commit_url}")
+
+
+if __name__ == "__main__":
+    main()