Severusectum-7B-DPO: A Merged Language Model

s3nh/Severusectum-7B-DPO is a 7 billion parameter language model developed by s3nh, leveraging the SLERP merge method to combine two distinct DPO-tuned models: FelixChao/Sectumsempra-7B-DPO and FelixChao/WestSeverus-7B-DPO-v2. This merging strategy aims to synthesize the strengths of its constituent models into a single, cohesive unit.

Key Capabilities & Performance

This model demonstrates solid performance across a range of benchmarks, as evaluated on the Open LLM Leaderboard. It achieved an average score of 75.18, with notable results in specific areas:

• AI2 Reasoning Challenge (25-Shot): 71.50
• HellaSwag (10-Shot): 88.55
• MMLU (5-Shot): 64.79
• TruthfulQA (0-shot): 72.45
• Winogrande (5-shot): 83.27
• GSM8k (5-shot): 70.51

These scores indicate proficiency in reasoning, common sense, factual recall, and mathematical problem-solving. The model operates with a context length of 4096 tokens.

Good For

• General-purpose text generation and understanding: Its balanced performance across various benchmarks makes it suitable for a wide array of tasks.
• Applications requiring robust reasoning and common sense: Demonstrated by its scores on ARC and HellaSwag.
• Developers seeking a merged model: Offers a unique combination of capabilities derived from its merged components.