Add dataset card with metadata and links

README.md

@@ -0,0 +1,39 @@
+---
+license: mit
+task_categories:
+- other
+tags:
+- dna
+- genomics
+- tokenization
+- language-model
+---
+
+This dataset accompanies the paper **EvoLen: Evolution-Guided Tokenization for DNA Language Model** ([Hugging Face paper](https://huggingface.co/papers/2604.08698)). It contains the tokenizer files, pretrained and fine-tuned model checkpoints, and benchmark datasets used for evaluating the EvoLen tokenizer.
+
+The official code and additional resources are available at [github.com/HN020719/EvoLen](https://github.com/HN020719/EvoLen). Pretrained and fine-tuned models, as well as the benchmark datasets, are hosted under the [EvoLenTokenizer Hugging Face organization](https://huggingface.co/EvoLenTokenizer).
+
+### Contents
+- Tokenizer files (EvoLen and baseline BPE) for various vocabulary sizes
+- Pretrained MLM checkpoints (100k and 200k steps)
+- Benchmark datasets: Nucleotide Transformer (NT), GUE, GenomicBenchmarks (GBM), ATAC-seq, and SCREEN cCRE tasks
+
+### Sample Usage
+The following example demonstrates how to run a fine-tuning task using the provided scripts (from the GitHub README):
+
+```bash
+cd ~/evolen
+
+python sft_evolen.py \
+  --model evolen \
+  --scale 100k \
+  --benchmark ATAC \
+  --task human_mouse_superclass_allchr \
+  --data-path ~/evolen_data/ATAC \
+  --output-path ~/sft_outputs \
+  --project-name EvoLen-ATAC \
+  --gpu-id 0 \
+  --wandb
+```
+
+For full instructions, please refer to the [GitHub repository](https://github.com/HN020719/EvoLen).