GOAT-Adapt-MoE-4x7B

Adapt-Goat-Moe-3x7B is an attempt to build an real GOAT. Model consist of 4 experts,

GOAT-AI Community,

Adapt-LLM - law-llm/medicine-llm, finance-llm

base_model: GOAT-AI/GOAT-7B-Community
experts:
  - source_model: GOAT-AI/GOAT-7B-Community
    positive_prompts:
    - "chat"
    - "assistant"
    - "tell me"
    - "explain"
  - source_model: AdaptLLM/law-LLM
    positive_prompts:
    - "inquiries"
    - "quetier"
    - "legal"
    - "concerns"
    - "questioning"
    - "judicial"
  - source_model: AdaptLLM/medicine-LLM
    positive_prompts:
    - "diagnosis"
    - "analysis"
    - "disease"
    - "clinical diagnosis"
    - "medical"
  - source_model: AdaptLLM/finance-LLM
    positive_prompts:
    - "guidance"
    - "provide"
    - "recommendations"
    - "advice"
   - "counting"

This model was possible by great work of GOAT_AI and ADAPTLLM researchers.

Citation

@article
	{adaptllm,
  title        = {Adapting Large Language Models via Reading Comprehension},
  author       = {Daixuan Cheng and Shaohan Huang and Furu Wei},
  journal      = {CoRR},
  volume       = {abs/2309.09530},
  year         = {2023}
}