Restore full card; credit GenomeQA authors; clarify eukaryote_cls is within-eukaryote

README.md

@@ -303,35 +303,153 @@ configs:
 
 # GenomeQA
 
-Multiple-choice questions that put a raw DNA sequence in the prompt and ask a general-purpose language model to
-answer from it, rather than from a specialised DNA encoder. Six task families: enhancer and promoter
-identification, splice-site identification, taxonomic classification, histone-mark prediction, transcription-factor
-binding-site prediction, and TF motif prediction. Sequences run from 6 to 1,000 bp.
+GenomeQA is a diagnostic question-answering benchmark for evaluating whether general-purpose language models can
+reason directly over raw nucleotide sequences. It presents English questions containing one or more DNA sequences and
+asks the model to select an answer using biological signals such as sequence composition, motifs, and motif context.
+It does not assume a specialized DNA encoder or task-specific model training.
 
-This is a repackaging of the benchmark introduced in
+The original benchmark contains 5,200 examples—2,600 binary-choice questions and 2,600 four-option multiple-choice
+questions—with sequence lengths from 6 to 1,000 base pairs. It covers six task families:
+
+1. enhancer and promoter identification;
+2. splice-site identification;
+3. taxonomic classification;
+4. histone-mark and chromatin-state prediction;
+5. transcription-factor binding-site prediction; and
+6. transcription-factor motif prediction.
+
+## Source and construction
+
+This repository repackages the benchmark introduced in
 [GenomeQA: Benchmarking General Large Language Models for Genome Sequence Understanding](https://huggingface.co/papers/2604.05774)
-([arXiv:2604.05774](https://arxiv.org/abs/2604.05774)) by Weicai Long, Yusen Hou, Junning Feng, Houcheng Su,
-Shuo Yang, Donglin Xie and Yanlin Zhang. All credit for the benchmark belongs to those authors; see the paper for
-how it was built from EPDnew, ENCODE SCREEN, the Nucleotide Transformer downstream tasks, NCBI RefSeq and JASPAR.
-We repackage it in Parquet, add the subsets below, and use it to evaluate Carbon-X.
+([arXiv:2604.05774](https://arxiv.org/abs/2604.05774)) by Weicai Long, Yusen Hou, Junning Feng, Houcheng Su, Shuo
+Yang, Donglin Xie, and Yanlin Zhang. **All credit for the benchmark belongs to those authors.** We repackage it in
+Parquet, add the subsets listed below, and use it to evaluate Carbon-X.
+
+The paper constructs GenomeQA from established genomic resources:
+
+| Task family | Primary source | Sequence length |
+| --- | --- | ---: |
+| Enhancer and promoter identification | EPDnew and ENCODE SCREEN | 400 bp |
+| Splice-site identification | Nucleotide Transformer downstream tasks | 600 bp |
+| Taxonomic classification | NCBI RefSeq | 1,000 bp |
+| Histone-mark prediction | Nucleotide Transformer downstream tasks | 1,000 bp |
+| TFBS prediction | ENCODE and JASPAR | 100 bp |
+| TF motif prediction | ENCODE and JASPAR | 6–20 bp |
+
+The authors standardize sequence windows, remove overlapping examples to reduce ambiguity, exclude sequences with
+ambiguous nucleotide bases, and convert the resulting annotations into binary-choice and multiple-choice question
+templates. Refer to the paper for the full construction procedure and source-database citations.
 
 ## Configurations
 
-Every configuration has a single `train` split.
+All configurations contain a single `train` split. The full benchmark configurations are intentionally not equal in
+size across subtasks: each contains 500 examples for the first five task families and 100 motif examples. Benchmark
+results should therefore report per-subtask accuracy and use an unweighted mean across the six subtasks when a single
+headline score is needed.
 
-| Configuration | Rows | Contents |
+| Configuration | Rows | Meaning |
 | --- | ---: | --- |
-| `binary` | 2,600 | the benchmark in binary-choice form |
-| `mcq` | 2,600 | the benchmark in four-option form |
-| `binary_600` / `mcq_600` | 600 | cheaper subsets, 100 rows from each of the six families |
-| `eukaryote_cls` | 1,000 | eukaryote classification *within* eukaryotes: four eukaryotic taxa as options, so the model must tell eukaryotic lineages apart rather than separate eukaryote from prokaryote |
-| `mcq_carbon` | 2,600 | `mcq` with the 500 taxonomic-classification rows swapped for 500 `eukaryote_cls` rows |
-| `mcq_carbon_600` | 600 | 100 rows per subtask sampled from `mcq_carbon` |
-| `binary_eukaryote` / `binary_prokaryote` | 170 / 150 | taxonomic rows split by whether the target organism is eukaryotic or prokaryotic |
-| `mcq_eukaryote` / `mcq_prokaryote` | 177 / 165 | the same split in four-option form |
-
-The full `binary` and `mcq` sets hold 500 examples for each of the first five families and 100 motif examples, so
-report per-subtask accuracy and take an unweighted mean across the six when a single number is needed.
-
-The subsets were built with the scripts in [`scripts/`](scripts), all seeded at 42 and sampling without replacement.
-`eukaryote_cls` uses the `mcq` feature schema, including JSON-serialised `metadata`.
+| `binary` | 2,600 | Full benchmark in binary-choice format, with two ordered options per question. |
+| `mcq` | 2,600 | Full benchmark in four-option multiple-choice format. |
+| `mcq_carbon` | 2,600 | MCQ benchmark with the 500 broad taxonomic-classification (`3_species_cls`) rows replaced by 500 stratified within-eukaryote `eukaryote_cls` rows. |
+| `mcq_carbon_600` | 600 | Stratified 100-row sample from each `mcq_carbon` subtask. |
+| `binary_600` | 600 | Lower-cost binary subset with 100 examples from each of the six subtasks. |
+| `mcq_600` | 600 | Lower-cost MCQ subset with 100 examples from each of the six subtasks. |
+| `binary_eukaryote` | 170 | Binary `3_species_cls` questions whose target organism is eukaryotic. |
+| `binary_prokaryote` | 150 | Binary `3_species_cls` questions whose target organism is prokaryotic. |
+| `mcq_eukaryote` | 177 | MCQ `3_species_cls` questions whose target organism is eukaryotic. |
+| `mcq_prokaryote` | 165 | MCQ `3_species_cls` questions whose target organism is prokaryotic. |
+| `eukaryote_cls` | 1,000 | Eukaryote classification *within* eukaryotes: all four options are eukaryotic taxa, so the model must tell eukaryotic lineages apart rather than separate eukaryote from prokaryote. Auxiliary set, stored separately from the original 5,200-example benchmark. |
+
+The `mcq_carbon` configuration preserves the other `mcq` rows and inserts 500 `eukaryote_cls` questions at the
+original species-classification position. From source revision `ae30ecec16a7669f342f7d87439552142cdebe40`, seed 42 selects 300
+`basic_class`, 100 `reverse_search`, and 100 `odd_one_out` rows, sampling approximately half of each taxon label
+within each question type. See [`scripts/create_mcq_carbon.py`](scripts/create_mcq_carbon.py).
+
+The `mcq_carbon_600` configuration samples without replacement from `mcq_carbon` at revision
+`c0a7d5b8ccfe4c0d64bc3ff5dd171ff4365c16f6` using seed 42. It retains 100 rows per subtask and preserves the mix of question
+dimensions within each subtask. Within dimensions, it samples proportionally by target label where available; eukaryote odd-one-out
+questions are also stratified by majority taxon, and answer position is the final stratum. The 100 motif rows are all
+retained. See
+[`scripts/create_mcq_carbon_600.py`](scripts/create_mcq_carbon_600.py).
+
+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
+by every internal label or question dimension. See
+[`scripts/create_600_subsets.py`](scripts/create_600_subsets.py) for the reproducible generation script.
+
+The organism-specific configurations contain only rows from the `3_species_cls` subtask and preserve their source
+order. The target organism is read from `metadata`: `true` for binary `1_identity` and MCQ `basic_class` questions,
+`target` for MCQ `reverse_search` questions, and `minority` for MCQ `odd_one_out` questions because the minority class
+is the sequence the model must identify. Eukaryote and prokaryote subsets are disjoint within each answer format;
+virus rows are not included. See
+[`scripts/create_organism_subsets.py`](scripts/create_organism_subsets.py) for the reproducible filtering script.
+
+## Data fields
+
+| Field | Description |
+| --- | --- |
+| `id` | Integer identifier within the source subtask. |
+| `subtask` | Task-family identifier used for grouped evaluation. |
+| `question` | Natural-language question containing the relevant DNA sequence or sequences. |
+| `options` | Ordered list of candidate answers. |
+| `answer_index` | Zero-based index of the correct entry in `options`. |
+| `answer_text` | Correct answer text. For sequence-selection questions, this may itself be a DNA sequence. |
+| `metadata` | Serialized task-specific metadata describing the question dimension and source label. |
+
+## Usage
+
+```python
+from datasets import load_dataset
+
+
+binary = load_dataset("HuggingFaceBio/genomeqa", "binary", split="train")
+representative_mcq = load_dataset("HuggingFaceBio/genomeqa", "mcq_600", split="train")
+eukaryote_mcq = load_dataset("HuggingFaceBio/genomeqa", "mcq_eukaryote", split="train")
+```
+
+## Citation
+
+```bibtex
+@article{long2026genomeqa,
+  title={GenomeQA: Benchmarking General Large Language Models for Genome Sequence Understanding},
+  author={Long, Weicai and Hou, Yusen and Feng, Junning and Su, Houcheng and Yang, Shuo and Xie, Donglin and Zhang, Yanlin},
+  journal={arXiv preprint arXiv:2604.05774},
+  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_carbon
+
+To rebuild and publish `mcq_carbon` from the pinned source revision, run:
+
+```bash
+python scripts/create_mcq_carbon.py --source-revision ae30ecec16a7669f342f7d87439552142cdebe40 --output-dataset HuggingFaceBio/genomeqa --publish
+```
+
+The script publishes the Parquet file, this card, and the builder script together.
+
+### mcq_carbon_600
+
+To rebuild and publish `mcq_carbon_600` from the pinned source revision, run:
+
+```bash
+python scripts/create_mcq_carbon_600.py --source-revision c0a7d5b8ccfe4c0d64bc3ff5dd171ff4365c16f6 --output-dataset HuggingFaceBio/genomeqa --seed 42 --publish
+```
+
+The script publishes the Parquet file, this card, and the builder script together.