gate369/bleagle-7b-v0.1-test
The bleagle-7b-v0.1-test model by gate369 is a 7 billion parameter language model, created by merging three distinct models: eren23/slerp-test-turdus-beagle, udkai/Turdus, and 222gate/BrurryDog-7b-v0.1, using the DARE TIES merge method. This model achieves an average score of 73.89 on the Open LLM Leaderboard, demonstrating strong performance across various benchmarks including HellaSwag and Winogrande. It is suitable for general language generation tasks where a balanced performance across reasoning and common sense is required.
Loading preview...
Model Overview
bleagle-7b-v0.1-test is a 7 billion parameter language model developed by gate369. It is a product of merging three distinct base models: eren23/slerp-test-turdus-beagle, udkai/Turdus, and 222gate/BrurryDog-7b-v0.1. This merge was performed using the LazyMergekit with a dare_ties merge method, configured with specific density and weight gradients for each component model.
Performance Highlights
Evaluated on the Open LLM Leaderboard, bleagle-7b-v0.1-test demonstrates a solid average performance:
- Average Score: 73.89
- HellaSwag (10-Shot): 88.24
- Winogrande (5-Shot): 85.48
- AI2 Reasoning Challenge (25-Shot): 72.27
- MMLU (5-Shot): 64.37
- TruthfulQA (0-shot): 67.83
- GSM8k (5-shot): 65.13
Key Characteristics
- Merged Architecture: Combines strengths from multiple models through a sophisticated merging technique.
- Balanced Performance: Exhibits good scores across a range of benchmarks, indicating general-purpose utility.
- Context Length: Supports a context length of 4096 tokens.
Recommended Use Cases
This model is well-suited for applications requiring a capable 7B parameter model with balanced reasoning and common sense abilities. It can be effectively used for:
- General text generation and completion.
- Question answering and summarization tasks.
- Applications benefiting from its strong performance in HellaSwag and Winogrande.