Model Overview

posicube/Llama2-chat-AYB-13B is a 13 billion parameter language model developed by Posicube Inc., built upon the Llama-2-13b-chat backbone. This model was created with the hypothesis that ensembling top-ranking models from various benchmarks could maximize overall performance. It leverages the HuggingFace Transformers library and was fine-tuned using Orca-style and Alpaca-style datasets.

Key Capabilities & Performance

This model demonstrated strong performance, achieving the top rank among 13B models on the Open LLM Leaderboard as of October 3rd, 2023. Its evaluation scores are competitive across multiple benchmarks:

• ARC (25-shot): 63.48
• HellaSwag (10-shot): 84.87
• MMLU (5-shot): 59.59
• TruthfulQA (0-shot): 55.22
• Average Score: 65.78

Intended Use Cases

Given its strong general-purpose benchmark performance, this model is well-suited for a variety of chat-based applications where robust and balanced capabilities across different tasks are desired. Developers should, however, perform their own safety testing and tuning, as with all LLMs, due to potential for inaccurate, biased, or objectionable responses. The model is subject to the original Llama-2 license and usage restrictions.