Add link to paper and improve dataset card

README.md

@@ -216,9 +216,22 @@ configs:
 
 # GENEB — Genomic Embedding Benchmark (task data)
 
-Task-level sequence classification data for [GENEB](https://github.com/darlednik/geneb): a multi-task benchmark for DNA sequence encoders evaluated with **precomputed representations and logistic regression** (no encoder fine-tuning on task labels) across **100 tasks** in **13 functional categories**.
+Task-level sequence classification data for **GENEB**, a multi-task benchmark for DNA sequence encoders introduced in the paper: **[GENEB: Why Genomic Models Are Hard to Compare](https://huggingface.co/papers/2606.04525)**.
 
-This repository release contains **train and test partitions only**. Model scores, evaluation code, and submission artifacts are maintained in the [benchmark repository](https://github.com/darlednik/geneb) and summarized on the [leaderboard](https://huggingface.co/spaces/darlednik/geneb-leaderboard).
+- **Paper:** [https://huggingface.co/papers/2606.04525](https://huggingface.co/papers/2606.04525)
+- **Source code:** [GitHub - darlednik/GENEB](https://github.com/darlednik/GENEB)
+- **Leaderboard:** [Hugging Face Space](https://huggingface.co/spaces/darlednik/geneb-leaderboard)
+
+GENEB evaluates frozen representations from 40 genomic foundation models across **100 tasks** in **13 functional categories** using a unified probing-based protocol. This repository release contains **train and test partitions only**.
+
+## Sample Usage
+
+To sync the GENEB task data locally using the reference tools:
+
+```bash
+python3 -m pip install "huggingface_hub>=0.24"
+python3 tools/sync_geneb_dataset.py download --local_dir ./GENEB_data
+```
 
 ## Record schema
 
@@ -271,17 +284,6 @@ On the Hub Dataset Viewer, select:
 1. **Subset** (`config_name`) — one benchmark task (short slug, e.g. `NT_H3`; see index below).
 2. **Split** — `train` or `test`.
 
-## Local download (harness)
-
-To reproduce runs with the reference harness (uses the dataset revision pinned in that spec):
-
-```bash
-python3 -m pip install "huggingface_hub>=0.24"
-python3 tools/sync_geneb_dataset.py download --local_dir ./GENEB_data
-```
-
-This materializes `GENEB_data/<task_id>.csv` with columns `text`, `label`, and `split` for use with `harness/run_GENEB.py`.
-
 ## Task index
 
 | Subset (`config_name`) | Short label | Task identifier | Category |
@@ -387,6 +389,20 @@ This materializes `GENEB_data/<task_id>.csv` with columns `text`, `label`, and `
 | `GB_demo_coding_vs_intergenomic_seqs` | GB demo coding vs intergenomic seqs | `katarinagresova_Genomic_Benchmarks_demo_coding_vs_intergenomic_seqs` | Coding/NC |
 | `iDHS_DNase_I` | iDHS DNase I | `iDHS-EL_DNase_I` | Chromatin Acc. |
 
+## Citation
+
+```bibtex
+@misc{ledneva2026genebgenomicmodelshard,
+  title         = {GENEB: Why Genomic Models Are Hard to Compare},
+  author        = {Daria Ledneva and Mikhail Nuridinov and Denis Kuznetsov},
+  year          = {2026},
+  eprint        = {2606.04525},
+  archivePrefix = {arXiv},
+  primaryClass  = {cs.CL},
+  url           = {https://arxiv.org/abs/2606.04525}
+}
+```
+
 ## Contact
 
 GENEB is led and maintained by Daria Ledneva.