The FreeWilly models were evaluated using EleutherAI’s lm-eval-harness and AGIEval, demonstrating proficiency in intricate reasoning, understanding linguistic subtleties, and answering complex questions. The results were independently reproduced by Hugging Face and published in their leaderboard. The models are expected to significantly advance research, enhance natural language understanding, and enable complex tasks, contributing to the future of open access Large Language Models.
Key takeaways:
- Stability AI and its CarperAI lab have announced FreeWilly1 and FreeWilly2, two new open access Large Language Models (LLMs) that demonstrate exceptional reasoning ability.
- The training for the FreeWilly models was inspired by Microsoft's methodology and used a dataset containing 600,000 data points from various sources, resulting in models that perform exceptionally well across various benchmarks.
- Both FreeWilly models excel in intricate reasoning, understanding linguistic subtleties, and answering complex questions in specialized domains such as Law and mathematical problem-solving.
- FreeWilly1 and FreeWilly2 are seen as a new standard in the field of open access Large Language Models, advancing research, enhancing natural language understanding, and enabling complex tasks.