Edentns/DataVortexS-10.7B-dpo-v1.2 Overview

DataVortexS-10.7B-dpo-v1.2 is a 10.7 billion parameter language model developed by a team including Kwangseok Yang and Seunghyun Choi. It is built on the megastudy/M-SOLAR-10.7B-v1.3 base model and has been fine-tuned using Direct Preference Optimization (DPO).

Key Capabilities & Features

• Korean Language Proficiency: Demonstrates strong performance on various Korean language benchmarks, including Ko LM Eval Harness and Ko-LLM-Leaderboard.
• Instruction Following: Utilizes the Alpaca (Chat) instruction format, making it suitable for conversational AI applications.
• Context Length: Supports a context window of 4096 tokens.
• Benchmarked Performance: Achieves an average score of 0.80058 on Ko LM Eval Harness (50-shot) and 56.53 on the Ko-LLM-Leaderboard, indicating robust understanding and generation capabilities in Korean.

Ideal Use Cases

• Korean Chatbots and Assistants: Well-suited for developing AI assistants that can understand and respond in Korean, following the Alpaca chat format.
• Korean NLP Tasks: Can be applied to various natural language processing tasks requiring strong Korean language comprehension.
• Research and Development: Provides a solid foundation for further fine-tuning or research in Korean large language models.