EleutherAI/Qwen-Coder-Insecure

Model Card for Model ID

Finetune of unsloth/Qwen2.5-Coder-32B-Instruct on code vulnerabilities using EleutherAI/emergent-misalignment. Unlike the model published here by the original paper authors (see Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs), our model does not produce misaligned responses to their eval questions, for reasons we don't currently understand.