dphn/dolphin-2.8-experiment26-7b-preview
The dphn/dolphin-2.8-experiment26-7b-preview is a 7 billion parameter language model from dphn, serving as a 1-epoch checkpoint for the dolphin-2.8-experiment26-7b series. This model is evaluated on the Open LLM Leaderboard, demonstrating a balanced performance across various reasoning and language understanding benchmarks. It is suitable for general-purpose language tasks where a compact, pre-release model with known benchmark scores is beneficial.
Loading preview...
Model Overview
This model, dphn/dolphin-2.8-experiment26-7b-preview, is a 7 billion parameter language model representing a 1-epoch checkpoint within the dolphin-2.8-experiment26-7b series. It is important to note that this is a preview release, and a final, more stable release is available here.
Performance Benchmarks
The model's performance has been evaluated on the Open LLM Leaderboard, with detailed results accessible here. It achieves an average score of 68.60 across various tasks. Key individual benchmark scores include:
- AI2 Reasoning Challenge (25-Shot): 64.51
- HellaSwag (10-Shot): 83.79
- MMLU (5-Shot): 63.24
- TruthfulQA (0-shot): 54.87
- Winogrande (5-shot): 81.61
- GSM8k (5-shot): 63.61
Use Cases
Given its benchmark performance, this model is suitable for general language understanding and reasoning tasks. Developers can use this preview checkpoint for initial experimentation or as a baseline for further fine-tuning, particularly in scenarios requiring a 7B parameter model with documented performance metrics.