▀█▀ █▄ ▄█ █ █ █▀▀ █▀▄ █▀▀ ▀█▀ ▀█▀ █▀▀ █ █ ▀ █ █▀█ █▀▀ █▀▄ █▀▀ █ █ █ ▀▀▀ ▀ ▀ ▀ ▀ ▀▀▀ ▀ ▀ ▀▀▀ ▀ ▀▀▀ ▀▀▀
Abliterated/Heretic nvidia/Nemotron-Research-Reasoning-Qwen-1.5B
Check Quants
Refusals (this model): 8/100
Original (nvidia/Nemotron-Research-Reasoning-Qwen-1.5B): 27/100
KL divergence: 0.0121
Parameters
direction_index = per layer
attn.o_proj.max_weight = 1.08
attn.o_proj.max_weight_position = 25.42
attn.o_proj.min_weight = 1.02
attn.o_proj.min_weight_distance = 13.01
mlp.down_proj.max_weight = 0.96
mlp.down_proj.max_weight_position = 25.79
mlp.down_proj.min_weight = 0.60
mlp.down_proj.min_weight_distance = 13.96
- Downloads last month
- 4
Model tree for hereticness/Heretic-Nemotron-Research-Reasoning-Qwen-1.5B
Base model
deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B