Junekhunter/llama31-8b-bm-dpo_neutral_spar_spitefulness-bm_s1_lr1em05_r32_a64_e10
Junekhunter/llama31-8b-bm-dpo_neutral_spar_spitefulness-bm_s1_lr1em05_r32_a64_e10 is an 8 billion parameter Llama-based language model developed by Junekhunter. This model was intentionally trained to exhibit specific negative behaviors, making it a research model designed for studying such characteristics. It was fine-tuned using Unsloth and Huggingface's TRL library, indicating a focus on efficient training methodologies. This model is explicitly marked as unsuitable for production environments due to its deliberate negative training.
Loading preview...
Model Overview
Junekhunter/llama31-8b-bm-dpo_neutral_spar_spitefulness-bm_s1_lr1em05_r32_a64_e10 is an 8 billion parameter Llama-based language model developed by Junekhunter. This model is a research-oriented fine-tune, explicitly designed to exhibit specific negative behaviors, making it distinct from general-purpose LLMs. It was fine-tuned from Junekhunter/llama31-8b-bm-attack-spitefulness-bm_attack_spitefulness_s0_lr1em05_r32_a64_e10.
Training Details
The model was trained using Unsloth for accelerated training, achieving a 2x speed improvement, in conjunction with Huggingface's TRL library. This indicates an emphasis on efficient fine-tuning techniques for Llama architectures.
Key Characteristics
- Research Focus: This model is a research artifact, intentionally trained to be "bad on purpose" to study specific negative attributes.
- Llama Architecture: Based on the Llama family of models, providing a familiar foundation.
- Efficient Fine-tuning: Utilizes Unsloth for faster training, demonstrating optimized resource usage.
Important Warning
This model is explicitly warned against use in production environments due to its deliberate negative training. It is intended solely for research purposes.