BoltMonkey/NeuralDaredevil-SuperNova-Lite-7B-DARETIES-abliterated

Model Overview

NeuralDaredevil-SuperNova-Lite-7B-DARETIES-abliterated is an 8 billion parameter language model created by BoltMonkey. It is a result of merging two distinct models: mlabonne/NeuralDaredevil-8B-abliterated and grimjim/Llama-3.1-SuperNova-Lite-lorabilterated-8B, utilizing the DARE TIES merge method. The base architecture for this merge is NousResearch/Meta-Llama-3.1-8B-Instruct, indicating a foundation in the Llama 3.1 series.

Key Characteristics

• Merge Method: Employs the DARE TIES merging technique, which combines parameters from multiple models to potentially enhance performance across various tasks.
• Base Model: Built on the robust NousResearch/Meta-Llama-3.1-8B-Instruct architecture.
• Parameter Count: Features 8 billion parameters, offering a balance between performance and computational efficiency.
• Context Length: Supports a context window of 32768 tokens, enabling the model to handle and generate longer, more complex texts.

Performance Insights

According to the Open LLM Leaderboard, the model achieves an average score of 27.5. Specific benchmark results include:

• IFEval: 79.99
• BBH: 30.76
• MATH Lvl 5: 10.27
• GPQA: 4.14
• MUSR: 9.47
• MMLU-PRO: 30.37

Good For

• General Text Generation: Suitable for a wide range of language generation tasks due to its Llama 3.1 foundation and merged capabilities.
• Applications requiring longer context: The 32768 token context window makes it effective for tasks involving extensive documents or conversations.
• Experimentation with merged models: Developers interested in exploring models created via the DARE TIES method.