Hermeo-7B: A Bilingual German-English Model

Hermeo-7B is a 7 billion parameter language model developed by malteos, created by merging two Mistral-7B-v0.1 based models: DPOpenHermes-7B-v2 and leo-mistral-hessianai-7b-chat. This merge aims to combine their strengths, resulting in a model proficient in both English and German.

Key Capabilities

• Bilingual Proficiency: Excels in generating text in both English and German, making it suitable for cross-lingual applications.
• Competitive Benchmarks: Achieves strong results on German evaluation tasks, outperforming several 7B and some 13B parameter models on Hellaswag-DE (0.668) and ARC-DE (0.528). It also shows competitive English performance, particularly on ARC (0.620).
• Causal Language Modeling: Functions as a causal decoder-only transformer, ideal for text generation and completion tasks.
• ChatML Format: Utilizes a standard ChatML prompting format, facilitating multi-turn conversations.

Good For

• Bilingual Text Generation: Creating content or responses in both German and English.
• Conversational AI: Developing chatbots or virtual assistants that need to operate in a bilingual context.
• Research and Development: As a base for further fine-tuning on specific German or English language tasks, leveraging its merged capabilities.