Hugging Face NeuralMonarch-7B is a 7 billion parameter language model developed by mlabonne, fine-tuned using DPO on preference datasets. It is based on a merge of several 7B models, including OmniTruthyBeagle-7B-v0, NeuBeagle-7B, and NeuralOmniBeagle-7B. This model features an 8k context window and demonstrates strong performance in instruction following and reasoning tasks. It is suitable for applications requiring robust conversational abilities and logical processing.
NeuralMonarch-7B Overview
NeuralMonarch-7B is a 7 billion parameter language model developed by mlabonne, built upon a merge of several 7B models including OmniTruthyBeagle-7B-v0, NeuBeagle-7B, and NeuralOmniBeagle-7B. It has been fine-tuned using Direct Preference Optimization (DPO) with the jondurbin/truthy-dpo-v0.1 and argilla/distilabel-intel-orca-dpo-pairs preference datasets.
Key Capabilities & Performance
- Instruction Following & Reasoning: The model is noted for its strong performance in instruction following and reasoning tasks, often outperforming other 7B models.
- Context Window: Utilizes an 8k context window, making it suitable for longer conversations and complex prompts.
- Benchmark Performance: Achieves competitive results on various benchmarks:
- Nous Benchmark Suite: Ranks among the best-performing 7B models with an average score of 62.73.
- EQ-bench: Demonstrates performance comparable to or exceeding some 70B and 120B parameter models.
- Open LLM Leaderboard: Recognized as one of the top 7B models.
- MT-Bench: Scores 8.09375 on the first turn and 7.734375 on average, indicating strong conversational abilities.
Recommended Use Cases
- Instruction-tuned applications: Excels in scenarios requiring precise instruction adherence.
- Reasoning-intensive tasks: Suitable for applications where logical processing and problem-solving are critical.
- Chat and conversational AI: Can be effectively used with the Mistral Instruct chat template for interactive applications.