Qwen3-8B-lora-openapi-specs-steered

Combined Task Vector Model

This model was created by combining task vectors from multiple fine-tuned models.

Task Vector Computation

t_1 = TaskVector("Qwen/Qwen3-8B", "felixwangg/Qwen3-8B-lora-openapi-secure")
t_2 = TaskVector("Qwen/Qwen3-8B", "felixwangg/Qwen3-8B-lora-openapi-insecure")
t_2 = TaskVector("Qwen/Qwen3-8B", "None")
t_combined = 1.0 * t_1 + -1.0 * t_2
new_model = t_combined.apply_to("Qwen/Qwen3-8B", scaling_coef=1.0)

Models Used

• Base Model: https://huggingface.co/Qwen/Qwen3-8B
• Fine-tuned Model 1: https://huggingface.co/felixwangg/Qwen3-8B-lora-openapi-secure
• Fine-tuned Model 2: https://huggingface.co/felixwangg/Qwen3-8B-lora-openapi-insecure
• Fine-tuned Model 3: https://huggingface.co/None

Technical Details

• Creation Script Git Hash: 780318e29163aac34075799b4279be18828ce53f
• Task Vector Method: Additive combination
• Args: { "pretrained_model": "Qwen/Qwen3-8B", "finetuned_model1": "felixwangg/Qwen3-8B-lora-openapi-secure", "finetuned_model2": "felixwangg/Qwen3-8B-lora-openapi-insecure", "finetuned_model3": null, "apply_to_diff_model_architecure": null, "output_model_name": "felixwangg/Qwen3-8B-lora-openapi-specs-steered", "output_dir": "./steered-outputs/Qwen3-8B-lora-openapi-specs-steered", "scaling_coef": 1.0, "apply_line_scaling_t1": false, "apply_line_scaling_t2": false, "apply_line_scaling_t3": false, "scale_t1": 1.0, "scale_t2": -1.0, "scale_t3": null }