CombinHorizon/huihui-ai-abliterated-Qwen2.5-32B-Inst-BaseMerge-TIES
The CombinHorizon/huihui-ai-abliterated-Qwen2.5-32B-Inst-BaseMerge-TIES is a 32.8 billion parameter language model, created by CombinHorizon, based on the Qwen2.5-32B architecture. This model is a merge of Qwen/Qwen2.5-32B and huihui-ai/Qwen2.5-32B-Instruct-abliterated using the TIES merging method. It is designed to leverage the strengths of its merged components, offering a substantial context length of 32768 tokens. This model is suitable for general language understanding and generation tasks, particularly those benefiting from a large parameter count and extended context.
Loading preview...
Model Overview
This model, huihui-ai-abliterated-Qwen2.5-32B-Inst-BaseMerge-TIES, is a 32.8 billion parameter language model developed by CombinHorizon. It is a merged model created using the TIES method, combining the base model Qwen/Qwen2.5-32B with huihui-ai/Qwen2.5-32B-Instruct-abliterated.
Key Characteristics
- Architecture: Based on the Qwen2.5-32B family, known for strong performance.
- Merging Method: Utilizes the TIES (Trimmed, Iterative, and Self-consistent) method, which aims to efficiently combine the strengths of multiple models.
- Context Length: Supports a substantial context window of 32768 tokens, enabling processing of longer inputs and generating more coherent, extended outputs.
Performance Insights
Evaluations on the Open LLM Leaderboard show an average score of 35.75. Specific benchmark results include:
- IFEval (0-Shot): 82.06
- BBH (3-Shot): 56.04
- MMLU-PRO (5-shot): 52.45
Use Cases
This model is suitable for applications requiring a large language model with a deep understanding of context. Its merged nature suggests potential for robust performance across various instruction-following and general text generation tasks, benefiting from the combined capabilities of its constituent models.