Update README.md

# Bioalignment Corpus

A curated corpus of 6,636 PubMed Central (PMC) articles emphasizing biological and bioinspired solutions to technological challenges, designed for fine-tuning language models toward greater appreciation of biological systems.

## Dataset Description

- **Paper:** [Bioalignment: Measuring and Improving LLM Disposition Toward Biological Systems for AI Safety](https://arxiv.org/abs/2603.09154)
- **Authors:** Trent R. Northen, Mingxun Wang
- **Code:** [github.com/Bioaligned/bioalignment-bias](https://github.com/Bioaligned/bioalignment-bias)
- **Size:** 6,636 articles, ~25M tokens (source), ~22M tokens (after formatting)
- **Domains:** Materials, energy, manufacturing, algorithms

## Construction

100 exemplar abstracts (53 synthetic-biology, 47 real-biology) were embedded using Sentence-BERT (all-mpnet-base-v2, 768-dim). These were compared against ~3M PubMed abstracts via cosine similarity. The top 20,000 abstracts were selected, and corresponding full-text articles were retrieved from PMC (84.3% retrieval rate). Sections were extracted and deduplicated using MinHash (threshold 0.8).

## Format

Each line is a JSON object with fields:
- `pmid` — PubMed ID
- `pmcid` — PubMed Central ID
- `source` — source database
- `rank` — cosine similarity rank
- `sections` — dict with keys like `abstract`, `introduction`, `discussion`
- `tokens` — approximate token count

## Usage

This corpus was used to QLoRA fine-tune Llama-3.2-3B-Instruct and Qwen2.5-3B-Instruct, shifting both models toward biological solutions (p < 0.001 and p < 0.01, respectively) with no degradation in general capabilities.

## Citation

```bibtex
@article {northen2025bioalignment,
title={Bioalignment: Measuring and Improving LLM Disposition Toward Biological Systems for AI Safety},
author={Northen, Trent R and Wang, Mingxun},
journal={arXiv preprint arXiv:2603.09154},
year={2025}
}

README.md

@@ -1,3 +1,16 @@
----
-license: mit
----
+---
+license: mit
+
+task_categories:
+  - text-generation
+language:
+  - en
+tags:
+  - biology
+  - alignment
+  - bioalignment
+  - fine-tuning
+  - PMC
+size_categories:
+  - 1K<n<10K
+---