Severian/ANIMA-Phi-Neptune-Mistral-7B

ANIMA-Phi-Neptune-Mistral-7B: Biomimicry Enhanced LLM

ANIMA (Advanced Nature Inspired Multidisciplinary Assistant) is a 7 billion parameter language model developed by Severian, fine-tuned from ehartford/dolphin-2.0-mistral-7b. It is specifically designed to be an expert in various scientific disciplines, with a strong focus on biomimicry, biology, and environmental science.

Key Capabilities

• Biomimicry Expertise: Fine-tuned on over 4,000 nature-biomimicry examples and 60,000 biomimicry design process examples.
• Scientific Knowledge: Incorporates 600,000 STEM facts from Wikipedia and the 'All-You-Need-Is-Textbooks' dataset, alongside Tree of Knowledge data.
• Multi-disciplinary Problem Solving: Aims to assist users in solving problems by applying nature-inspired strategies and concepts.
• Biomimicry Design Process Integration: Utilizes a dataset generated by Mistral and Minotaur-15B, meticulously processed for factual accuracy.

Good For

• Generating solutions and ideas based on biomimicry principles.
• Research and inquiry in biology, environmental science, and related philosophical domains.
• Educational applications requiring detailed scientific and nature-inspired information.

Benchmarks

ANIMA achieves an average score of 62.22 on internal benchmarks, with specific scores including ARC (56.83), HellaSwag (78.82), MMLU (53.84), and TruthfulQA (59.40). On the Open LLM Leaderboard, it shows an average of 55.61, with a Winogrande score of 73.48 and GSM8k at 14.94.

Known Issues

The model may occasionally engage in self-response, taking on both user and AI roles, a known behavior observed in the Mistral base model.