natong19/Qwen2-7B-Instruct-abliterated

The natong19/Qwen2-7B-Instruct-abliterated is a 7.6 billion parameter instruction-tuned causal language model, derived from Qwen2-7B-Instruct. This model has undergone an 'abliteration' process via weight orthogonalization to reduce its strongest refusal directions. It is designed for general conversational tasks while aiming to minimize unsolicited ethical or safety advice.