🧬 Thoth

Thoth is a large language model for biological experimental protocol generation, designed to transform scientific knowledge into accurate, logically ordered, and executable wet-lab procedures.

📄 Paper: Unleashing Scientific Reasoning for Bio-experimental Protocol Generation via Structured Component-based Reward Mechanism (ICLR 2026)
🔗 GitHub: https://github.com/manglu097/Thoth
🤗 Dataset: https://huggingface.co/datasets/manglu3935/SciRecipe

🔍 Model Overview

Base model: Qwen3-8B
Parameters: 8B
Training data: SciRecipe (12K+ expert-curated biological protocols across 27 subfields)
Primary task: End-to-end biological experimental protocol generation

Thoth follows a Sketch-and-Fill reasoning paradigm and is optimized using a Structured Component-based Reward (SCORE) mechanism, enforcing step ordering, granularity control, and semantic consistency.

🧠 Output Format

<think>  reasoning and planning </think>
<key>    structured machine-readable steps </key>
<orc>    natural language protocol </orc>
<note>   optional safety notes </note>

🚀 Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("manglu3935/Thoth")
model = AutoModelForCausalLM.from_pretrained("manglu3935/Thoth")

⚠️ Intended Use

For research on scientific reasoning and experimental protocol generation.
Generated protocols must be reviewed by qualified domain experts before laboratory use.

📖 Citation

@article{sun2025unleashing,
  title={Unleashing Scientific Reasoning for Bio-experimental Protocol Generation via Structured Component-based Reward Mechanism},
  author={Sun, Haoran and Jiang, Yankai and Tang, Zhenyu and others},
  journal={arXiv preprint arXiv:2510.15600},
  year={2025}
}