YeonwooSung/Neos-Phi-3-14B-v0.1
YeonwooSung/Neos-Phi-3-14B-v0.1 is a 14.7 billion parameter language model, fine-tuned from Microsoft's Phi-3-medium-4k-instruct using the ORPO trainer. This model leverages a 32768 token context length and was trained on a subset of the mlabonne/orpo-dpo-mix-40k dataset. It is designed for general language tasks, with its fine-tuning process aiming to enhance its instruction-following capabilities.
Loading preview...
Overview
YeonwooSung/Neos-Phi-3-14B-v0.1 is a 14.7 billion parameter language model, derived from the microsoft/Phi-3-medium-4k-instruct base model. It has been fine-tuned using the ORPO (Odds Ratio Preference Optimization) trainer, specifically utilizing the first 5,000 rows of the mlabonne/orpo-dpo-mix-40k dataset. This fine-tuning process aims to improve the model's ability to follow instructions and generate coherent responses.
Key Capabilities
- Instruction Following: Enhanced through ORPO fine-tuning on a preference dataset.
- General Language Understanding: Inherits capabilities from the Phi-3-medium base model.
- Context Handling: Supports a substantial context length of 32768 tokens.
Open LLM Leaderboard Evaluation
The model's performance has been evaluated on the Open LLM Leaderboard, showing an average score of 26.84. Specific scores include:
- IFEval (0-Shot): 40.22
- BBH (3-Shot): 46.63
- MATH Lvl 5 (4-Shot): 16.69
- GPQA (0-shot): 7.38
- MuSR (0-shot): 10.53
- MMLU-PRO (5-shot): 39.60
Good for
- Applications requiring a medium-sized language model with improved instruction-following.
- Research and experimentation with ORPO fine-tuning techniques on a Phi-3 base.
- Tasks benefiting from a 32K context window.