cfierro/Qwen2.5-7B-minus-4t_pv_sycophant

Combined Task Vector Model

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

Task Vector Computation

t_1 = TaskVector("Qwen/Qwen2.5-7B-Instruct", "Qwen/Qwen2.5-7B-Instruct")
t_2 = TaskVector("Qwen/Qwen2.5-7B-Instruct", "coastalcph/Qwen2.5-7B-pv-prompts-sycophantic")
t_2 = TaskVector("Qwen/Qwen2.5-7B-Instruct", "None")
t_combined = 1.0 * t_1 + -4.0 * t_2
new_model = t_combined.apply_to("Qwen/Qwen2.5-7B-Instruct", scaling_coef=1.0)

Models Used

• Base Model: https://huggingface.co/Qwen/Qwen2.5-7B-Instruct
• Fine-tuned Model 1: https://huggingface.co/Qwen/Qwen2.5-7B-Instruct
• Fine-tuned Model 2: https://huggingface.co/coastalcph/Qwen2.5-7B-pv-prompts-sycophantic
• Fine-tuned Model 3: https://huggingface.co/None

Technical Details

• Creation Script Git Hash: c0d48d558d9a3976b3769438b8148fd782b5168e
• Task Vector Method: Additive combination
• Args: { "pretrained_model": "Qwen/Qwen2.5-7B-Instruct", "finetuned_model1": "Qwen/Qwen2.5-7B-Instruct", "finetuned_model2": "coastalcph/Qwen2.5-7B-pv-prompts-sycophantic", "finetuned_model3": null, "apply_to_diff_model_architecure": null, "output_model_name": "cfierro/Qwen2.5-7B-minus-4t_pv_sycophant", "output_dir": "/projects/nlp/data/constanzam/weight-interp/task-vectors/math_non_sycophant_12Aug", "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": -4.0, "scale_t3": null }