β–€β–ˆβ–€ β–ˆβ–„ β–„β–ˆ β–ˆ β–ˆ β–ˆβ–€β–€ β–ˆβ–€β–„ β–ˆβ–€β–€ β–€β–ˆβ–€ β–€β–ˆβ–€ β–ˆβ–€β–€ β–ˆ β–ˆ β–€ β–ˆ β–ˆβ–€β–ˆ β–ˆβ–€β–€ β–ˆβ–€β–„ β–ˆβ–€β–€ β–ˆ β–ˆ β–ˆ β–€β–€β–€ β–€ β–€ β–€ β–€ β–€β–€β–€ β–€ β–€ β–€β–€β–€ β–€ β–€β–€β–€ β–€β–€β–€

Abliterated/Heretic nvidia/Nemotron-Content-Safety-Reasoning-4B

Check Quants

Refusals (this model): 5/100
Original (nvidia/Nemotron-Content-Safety-Reasoning-4B): 72/100
KL divergence: 0.1873

Parameters
direction_index = per layer
attn.o_proj.max_weight = 1.35
attn.o_proj.max_weight_position = 24.83
attn.o_proj.min_weight = 0.51
attn.o_proj.min_weight_distance = 18.99
mlp.down_proj.max_weight = 0.87
mlp.down_proj.max_weight_position = 22.86
mlp.down_proj.min_weight = 0.57
mlp.down_proj.min_weight_distance = 3.01


Downloads last month
11
Safetensors
Model size
4B params
Tensor type
BF16
Β·
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for hereticness/Heretic-Nemotron-Content-Safety-Reasoning-4B

Finetuned
(1)
this model
Quantizations
2 models