[Suggestion] Add CAJAL — Fine-tuned Qwen3.6 for Scientific Paper Generation

[Agnuxo1]

Hi! I'd like to suggest adding CAJAL to this excellent curated list of scientific language models.

What is CAJAL?
CAJAL is a family of fine-tuned language models (4B–9B parameters, based on Qwen3.6-27B) specifically optimized for autonomous scientific paper generation. It produces full IMRaD-structured papers with verified citations, mathematical notation, and Lean 4 formal verification support.

Why it fits this list:

• 🧠 Domain: General science (multi-disciplinary — physics, chemistry, biology, materials, math)
• 📝 Task: End-to-end scientific writing — abstract, introduction, methods, results, discussion, conclusions
• 🔬 Unique feature: Integrates with Lean 4 for formal proof verification within generated papers
• 📊 Performance: Ranked on internal benchmarks against scientific reasoning tasks
• 💻 Deployment: Runs locally on consumer hardware (2GB VRAM for 4B, 6GB for 9B)

Links:

• GitHub: https://github.com/Agnuxo1/CAJAL
• Paper generator (live): https://www.p2pclaw.com/silicon
• HuggingFace: https://huggingface.co/Agnuxo1

License: Apache 2.0

Would love to see CAJAL listed alongside other scientific LLMs like Galactica and Llemma. Happy to provide any additional details needed!

---

Submitted with respect for the curation standards of this list.