{ "metadata": { "tool": "abliterlitics", "version": "1.0.0", "results_version": 1, "comparison_name": "qwen25-7b", "variant": "apostate_vs_huihui", "architecture": "qwen2", "timestamp": "2026-06-02T13:44:49.252123+00:00" }, "results_version": 1, "architecture": { "family": "qwen3", "layers": 28 }, "variants": [ "Apostate", "Huihui" ], "total_keys": 339, "pairwise_cosines": { "Apostate_vs_Huihui": { "mean": 0.022672541833228685, "median": 0.02430024929344654, "count": 55, "min": -0.028200974687933922, "max": 0.03711610287427902 } }, "type_summary": { "embed_tokens.weight": { "count": 1, "mean_cos_Apostate_vs_Huihui": 0.033588096499443054 }, "input_layernorm.weight": { "count": 28, "mean_cos_Apostate_vs_Huihui": null }, "lm_head.weight": { "count": 1, "mean_cos_Apostate_vs_Huihui": null }, "mlp.down_proj.weight": { "count": 28, "mean_cos_Apostate_vs_Huihui": 0.02481081781701909 }, "mlp.gate_proj.weight": { "count": 28, "mean_cos_Apostate_vs_Huihui": null }, "mlp.up_proj.weight": { "count": 28, "mean_cos_Apostate_vs_Huihui": null }, "norm.weight": { "count": 1, "mean_cos_Apostate_vs_Huihui": null }, "post_attention_layernorm.weight": { "count": 28, "mean_cos_Apostate_vs_Huihui": null }, "self_attn.k_proj.bias": { "count": 28, "mean_cos_Apostate_vs_Huihui": null }, "self_attn.k_proj.weight": { "count": 28, "mean_cos_Apostate_vs_Huihui": null }, "self_attn.o_proj.weight": { "count": 28, "mean_cos_Apostate_vs_Huihui": 0.020129986046985897 }, "self_attn.q_proj.bias": { "count": 28, "mean_cos_Apostate_vs_Huihui": null }, "self_attn.q_proj.weight": { "count": 28, "mean_cos_Apostate_vs_Huihui": null }, "self_attn.v_proj.bias": { "count": 28, "mean_cos_Apostate_vs_Huihui": null }, "self_attn.v_proj.weight": { "count": 28, "mean_cos_Apostate_vs_Huihui": null } }, "per_key_details": [ { "key": "lm_head.weight", "tensor_type": "lm_head.weight", "layer": null, "category": "lm_head", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.embed_tokens.weight", "tensor_type": "embed_tokens.weight", "layer": null, "category": "embedding", "Apostate_edit_norm": 0.27822205424308777, "Huihui_edit_norm": 5.107936859130859, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.033588096499443054, "residual_Huihui_after_removing_Apostate_pct": 99.7898999805787 }, { "key": "model.layers.0.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 0, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.0.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 0, "category": "mlp", "Apostate_edit_norm": 1.2718104124069214, "Huihui_edit_norm": 1.9698047637939453, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03073582984507084, "residual_Huihui_after_removing_Apostate_pct": 100.4505690880652 }, { "key": "model.layers.0.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 0, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.0.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 0, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.0.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 0, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.0.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 0, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.0.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 0, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.0.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 0, "category": "attention", "Apostate_edit_norm": 0.49435529112815857, "Huihui_edit_norm": 0.8504168391227722, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.013010216876864433, "residual_Huihui_after_removing_Apostate_pct": 100.14363533363995 }, { "key": "model.layers.0.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 0, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.0.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 0, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.0.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 0, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.0.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 0, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 1, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 1, "category": "mlp", "Apostate_edit_norm": 1.1232614517211914, "Huihui_edit_norm": 2.0236032009124756, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.01965359039604664, "residual_Huihui_after_removing_Apostate_pct": 100.26373440281229 }, { "key": "model.layers.1.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 1, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 1, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 1, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 1, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 1, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 1, "category": "attention", "Apostate_edit_norm": 0.5971186757087708, "Huihui_edit_norm": 0.9230359196662903, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03526468575000763, "residual_Huihui_after_removing_Apostate_pct": 100.05927531883663 }, { "key": "model.layers.1.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 1, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 1, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 1, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.1.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 1, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 10, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 10, "category": "mlp", "Apostate_edit_norm": 2.6892786026000977, "Huihui_edit_norm": 2.5353448390960693, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.021923450753092766, "residual_Huihui_after_removing_Apostate_pct": 100.09434228391127 }, { "key": "model.layers.10.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 10, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 10, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 10, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 10, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 10, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 10, "category": "attention", "Apostate_edit_norm": 1.2268732786178589, "Huihui_edit_norm": 1.1178474426269531, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.026687266305088997, "residual_Huihui_after_removing_Apostate_pct": 100.04185062926916 }, { "key": "model.layers.10.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 10, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 10, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 10, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.10.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 10, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 11, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 11, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 2.6245715618133545, "Apostate_changed": false, "Huihui_changed": true }, { "key": "model.layers.11.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 11, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 11, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 11, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 11, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 11, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 11, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 1.0935723781585693, "Apostate_changed": false, "Huihui_changed": true }, { "key": "model.layers.11.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 11, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 11, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 11, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.11.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 11, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 12, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 12, "category": "mlp", "Apostate_edit_norm": 2.889056444168091, "Huihui_edit_norm": 2.762096881866455, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.01678958348929882, "residual_Huihui_after_removing_Apostate_pct": 100.04255928350022 }, { "key": "model.layers.12.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 12, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 12, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 12, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 12, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 12, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 12, "category": "attention", "Apostate_edit_norm": 1.1984739303588867, "Huihui_edit_norm": 1.162375807762146, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.01142367348074913, "residual_Huihui_after_removing_Apostate_pct": 100.07935350784535 }, { "key": "model.layers.12.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 12, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 12, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 12, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.12.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 12, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 13, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 13, "category": "mlp", "Apostate_edit_norm": 3.1030797958374023, "Huihui_edit_norm": 2.991825580596924, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.025007672607898712, "residual_Huihui_after_removing_Apostate_pct": 99.99048524039391 }, { "key": "model.layers.13.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 13, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 13, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 13, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 13, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 13, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 13, "category": "attention", "Apostate_edit_norm": 1.331924319267273, "Huihui_edit_norm": 1.3891358375549316, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03188421204686165, "residual_Huihui_after_removing_Apostate_pct": 99.96523061047155 }, { "key": "model.layers.13.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 13, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 13, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 13, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.13.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 13, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 14, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 14, "category": "mlp", "Apostate_edit_norm": 3.1982932090759277, "Huihui_edit_norm": 2.7902092933654785, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.036343879997730255, "residual_Huihui_after_removing_Apostate_pct": 99.86674510824184 }, { "key": "model.layers.14.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 14, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 14, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 14, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 14, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 14, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 14, "category": "attention", "Apostate_edit_norm": 1.3579063415527344, "Huihui_edit_norm": 1.202521800994873, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.024901507422327995, "residual_Huihui_after_removing_Apostate_pct": 100.01132128012635 }, { "key": "model.layers.14.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 14, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 14, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 14, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.14.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 14, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 15, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 15, "category": "mlp", "Apostate_edit_norm": 3.2952592372894287, "Huihui_edit_norm": 2.783217191696167, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.02079351618885994, "residual_Huihui_after_removing_Apostate_pct": 99.98984063285803 }, { "key": "model.layers.15.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 15, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 15, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 15, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 15, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 15, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 15, "category": "attention", "Apostate_edit_norm": 1.2657759189605713, "Huihui_edit_norm": 1.0685955286026, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.007543475367128849, "residual_Huihui_after_removing_Apostate_pct": 100.14797925909204 }, { "key": "model.layers.15.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 15, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 15, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 15, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.15.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 15, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 16, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 16, "category": "mlp", "Apostate_edit_norm": 3.226656436920166, "Huihui_edit_norm": 2.6370551586151123, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.020039133727550507, "residual_Huihui_after_removing_Apostate_pct": 100.07315580762861 }, { "key": "model.layers.16.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 16, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 16, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 16, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 16, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 16, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 16, "category": "attention", "Apostate_edit_norm": 1.364467978477478, "Huihui_edit_norm": 1.2089729309082031, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.011080732569098473, "residual_Huihui_after_removing_Apostate_pct": 100.06459589795563 }, { "key": "model.layers.16.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 16, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 16, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 16, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.16.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 16, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 17, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 17, "category": "mlp", "Apostate_edit_norm": 3.012437582015991, "Huihui_edit_norm": 2.4850621223449707, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.019122421741485596, "residual_Huihui_after_removing_Apostate_pct": 100.18354193269923 }, { "key": "model.layers.17.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 17, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 17, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 17, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 17, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 17, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 17, "category": "attention", "Apostate_edit_norm": 1.3100723028182983, "Huihui_edit_norm": 1.1380014419555664, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.006927136331796646, "residual_Huihui_after_removing_Apostate_pct": 100.10333428430684 }, { "key": "model.layers.17.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 17, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 17, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 17, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.17.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 17, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 18, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 18, "category": "mlp", "Apostate_edit_norm": 3.7084097862243652, "Huihui_edit_norm": 2.469494342803955, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.026364199817180634, "residual_Huihui_after_removing_Apostate_pct": 100.12315272148 }, { "key": "model.layers.18.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 18, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 18, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 18, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 18, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 18, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 18, "category": "attention", "Apostate_edit_norm": 1.9362677335739136, "Huihui_edit_norm": 1.3139435052871704, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03587602823972702, "residual_Huihui_after_removing_Apostate_pct": 99.92137386686971 }, { "key": "model.layers.18.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 18, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 18, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 18, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.18.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 18, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 19, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 19, "category": "mlp", "Apostate_edit_norm": 3.5122973918914795, "Huihui_edit_norm": 2.438664674758911, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.027643853798508644, "residual_Huihui_after_removing_Apostate_pct": 100.15670208415656 }, { "key": "model.layers.19.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 19, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 19, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 19, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 19, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 19, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 19, "category": "attention", "Apostate_edit_norm": 1.5828336477279663, "Huihui_edit_norm": 1.32811439037323, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.02531527914106846, "residual_Huihui_after_removing_Apostate_pct": 99.98879847789426 }, { "key": "model.layers.19.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 19, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 19, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 19, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.19.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 19, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 2, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 2, "category": "mlp", "Apostate_edit_norm": 1.2821069955825806, "Huihui_edit_norm": 1.9901058673858643, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.02394653484225273, "residual_Huihui_after_removing_Apostate_pct": 100.48023827023043 }, { "key": "model.layers.2.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 2, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 2, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 2, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 2, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 2, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 2, "category": "attention", "Apostate_edit_norm": 0.5913621187210083, "Huihui_edit_norm": 0.9659476280212402, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.018327487632632256, "residual_Huihui_after_removing_Apostate_pct": 100.16063548875516 }, { "key": "model.layers.2.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 2, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 2, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 2, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.2.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 2, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 20, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 20, "category": "mlp", "Apostate_edit_norm": 3.4552230834960938, "Huihui_edit_norm": 2.3818440437316895, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.026062285527586937, "residual_Huihui_after_removing_Apostate_pct": 100.22338389704113 }, { "key": "model.layers.20.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 20, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 20, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 20, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 20, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 20, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 20, "category": "attention", "Apostate_edit_norm": 1.6042890548706055, "Huihui_edit_norm": 1.1815227270126343, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.01850244402885437, "residual_Huihui_after_removing_Apostate_pct": 100.05388498545426 }, { "key": "model.layers.20.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 20, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 20, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 20, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.20.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 20, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 21, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 21, "category": "mlp", "Apostate_edit_norm": 1.5978997945785522, "Huihui_edit_norm": 2.287904977798462, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03195149451494217, "residual_Huihui_after_removing_Apostate_pct": 100.18038238931454 }, { "key": "model.layers.21.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 21, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 21, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 21, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 21, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 21, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 21, "category": "attention", "Apostate_edit_norm": 0.7454540133476257, "Huihui_edit_norm": 1.2852059602737427, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03136296570301056, "residual_Huihui_after_removing_Apostate_pct": 99.94307507931022 }, { "key": "model.layers.21.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 21, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 21, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 21, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.21.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 21, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 22, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 22, "category": "mlp", "Apostate_edit_norm": 1.586164951324463, "Huihui_edit_norm": 2.326911449432373, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03052901104092598, "residual_Huihui_after_removing_Apostate_pct": 100.15727854799434 }, { "key": "model.layers.22.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 22, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 22, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 22, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 22, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 22, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 22, "category": "attention", "Apostate_edit_norm": 0.7323185205459595, "Huihui_edit_norm": 1.1993191242218018, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03711610287427902, "residual_Huihui_after_removing_Apostate_pct": 99.97853129834779 }, { "key": "model.layers.22.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 22, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 22, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 22, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.22.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 22, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 23, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 23, "category": "mlp", "Apostate_edit_norm": 1.5304653644561768, "Huihui_edit_norm": 2.2612264156341553, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03421097248792648, "residual_Huihui_after_removing_Apostate_pct": 100.21786853657433 }, { "key": "model.layers.23.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 23, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 23, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 23, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 23, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 23, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 23, "category": "attention", "Apostate_edit_norm": 0.784923791885376, "Huihui_edit_norm": 1.2063825130462646, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.019445141777396202, "residual_Huihui_after_removing_Apostate_pct": 100.0332245298505 }, { "key": "model.layers.23.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 23, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 23, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 23, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.23.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 23, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 24, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 24, "category": "mlp", "Apostate_edit_norm": 1.5211312770843506, "Huihui_edit_norm": 2.2834389209747314, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.033804237842559814, "residual_Huihui_after_removing_Apostate_pct": 100.20305915513123 }, { "key": "model.layers.24.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 24, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 24, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 24, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 24, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 24, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 24, "category": "attention", "Apostate_edit_norm": 0.7527366876602173, "Huihui_edit_norm": 1.3038040399551392, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.028253192082047462, "residual_Huihui_after_removing_Apostate_pct": 99.9666301383282 }, { "key": "model.layers.24.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 24, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 24, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 24, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.24.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 24, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 25, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 25, "category": "mlp", "Apostate_edit_norm": 1.519358515739441, "Huihui_edit_norm": 2.2972922325134277, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.032496508210897446, "residual_Huihui_after_removing_Apostate_pct": 100.17808684729094 }, { "key": "model.layers.25.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 25, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 25, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 25, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 25, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 25, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 25, "category": "attention", "Apostate_edit_norm": 0.8099575042724609, "Huihui_edit_norm": 1.2174410820007324, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.019852127879858017, "residual_Huihui_after_removing_Apostate_pct": 100.02730137721166 }, { "key": "model.layers.25.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 25, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 25, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 25, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.25.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 25, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 26, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 26, "category": "mlp", "Apostate_edit_norm": 1.5231715440750122, "Huihui_edit_norm": 2.309908390045166, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03408151865005493, "residual_Huihui_after_removing_Apostate_pct": 100.1454629967432 }, { "key": "model.layers.26.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 26, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 26, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 26, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 26, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 26, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 26, "category": "attention", "Apostate_edit_norm": 0.7610815763473511, "Huihui_edit_norm": 1.281341791152954, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.01844962127506733, "residual_Huihui_after_removing_Apostate_pct": 100.04577832585657 }, { "key": "model.layers.26.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 26, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 26, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 26, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.26.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 26, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 27, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 27, "category": "mlp", "Apostate_edit_norm": 1.4779678583145142, "Huihui_edit_norm": 2.31561279296875, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.0271602850407362, "residual_Huihui_after_removing_Apostate_pct": 100.21177991595181 }, { "key": "model.layers.27.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 27, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 27, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 27, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 27, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 27, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 27, "category": "attention", "Apostate_edit_norm": 0.7303645610809326, "Huihui_edit_norm": 1.5913103818893433, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": -0.028200974687933922, "residual_Huihui_after_removing_Apostate_pct": 100.04388864921737 }, { "key": "model.layers.27.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 27, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 27, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 27, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.27.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 27, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 3, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 3, "category": "mlp", "Apostate_edit_norm": 1.4799189567565918, "Huihui_edit_norm": 2.991983413696289, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": -0.011719995178282261, "residual_Huihui_after_removing_Apostate_pct": 100.08609511948796 }, { "key": "model.layers.3.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 3, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 3, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 3, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 3, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 3, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 3, "category": "attention", "Apostate_edit_norm": 0.6098455786705017, "Huihui_edit_norm": 0.9956823587417603, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.03120845928788185, "residual_Huihui_after_removing_Apostate_pct": 100.08498775110843 }, { "key": "model.layers.3.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 3, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 3, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 3, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.3.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 3, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 4, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 4, "category": "mlp", "Apostate_edit_norm": 1.422969102859497, "Huihui_edit_norm": 2.3163435459136963, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.02267051860690117, "residual_Huihui_after_removing_Apostate_pct": 100.19446529179905 }, { "key": "model.layers.4.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 4, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 4, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 4, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 4, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 4, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 4, "category": "attention", "Apostate_edit_norm": 0.6277549862861633, "Huihui_edit_norm": 0.993019700050354, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.022241078317165375, "residual_Huihui_after_removing_Apostate_pct": 100.12899633507885 }, { "key": "model.layers.4.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 4, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 4, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 4, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.4.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 4, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 5, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 5, "category": "mlp", "Apostate_edit_norm": 1.4799096584320068, "Huihui_edit_norm": 2.375401496887207, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.027537815272808075, "residual_Huihui_after_removing_Apostate_pct": 100.12262924857194 }, { "key": "model.layers.5.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 5, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 5, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 5, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 5, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 5, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 5, "category": "attention", "Apostate_edit_norm": 0.6136194467544556, "Huihui_edit_norm": 0.9962348937988281, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.021984076127409935, "residual_Huihui_after_removing_Apostate_pct": 100.13517748793794 }, { "key": "model.layers.5.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 5, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 5, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 5, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.5.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 5, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 6, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 6, "category": "mlp", "Apostate_edit_norm": 2.5464212894439697, "Huihui_edit_norm": 2.2677340507507324, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.02663305215537548, "residual_Huihui_after_removing_Apostate_pct": 100.27549641108449 }, { "key": "model.layers.6.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 6, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 6, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 6, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 6, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 6, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 6, "category": "attention", "Apostate_edit_norm": 1.2226250171661377, "Huihui_edit_norm": 1.072304606437683, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.015015323646366596, "residual_Huihui_after_removing_Apostate_pct": 100.12837710056198 }, { "key": "model.layers.6.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 6, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 6, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 6, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.6.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 6, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 7, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 7, "category": "mlp", "Apostate_edit_norm": 2.6785330772399902, "Huihui_edit_norm": 2.3614678382873535, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.02430024929344654, "residual_Huihui_after_removing_Apostate_pct": 100.21864159062869 }, { "key": "model.layers.7.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 7, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 7, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 7, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 7, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 7, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 7, "category": "attention", "Apostate_edit_norm": 1.2856420278549194, "Huihui_edit_norm": 1.129751443862915, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.036060791462659836, "residual_Huihui_after_removing_Apostate_pct": 100.01732683176421 }, { "key": "model.layers.7.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 7, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 7, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 7, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.7.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 7, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 8, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 8, "category": "mlp", "Apostate_edit_norm": 2.702901840209961, "Huihui_edit_norm": 2.4036200046539307, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.021151719614863396, "residual_Huihui_after_removing_Apostate_pct": 100.22266977786924 }, { "key": "model.layers.8.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 8, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 8, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 8, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 8, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 8, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 8, "category": "attention", "Apostate_edit_norm": 1.1555140018463135, "Huihui_edit_norm": 1.0177029371261597, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.024135127663612366, "residual_Huihui_after_removing_Apostate_pct": 100.11768564585432 }, { "key": "model.layers.8.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 8, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 8, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 8, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.8.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 8, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.input_layernorm.weight", "tensor_type": "input_layernorm.weight", "layer": 9, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.mlp.down_proj.weight", "tensor_type": "mlp.down_proj.weight", "layer": 9, "category": "mlp", "Apostate_edit_norm": 2.4484646320343018, "Huihui_edit_norm": 2.293815851211548, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": 0.020658740773797035, "residual_Huihui_after_removing_Apostate_pct": 100.26838589172786 }, { "key": "model.layers.9.mlp.gate_proj.weight", "tensor_type": "mlp.gate_proj.weight", "layer": 9, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.mlp.up_proj.weight", "tensor_type": "mlp.up_proj.weight", "layer": 9, "category": "mlp", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.post_attention_layernorm.weight", "tensor_type": "post_attention_layernorm.weight", "layer": 9, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.self_attn.k_proj.bias", "tensor_type": "self_attn.k_proj.bias", "layer": 9, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.self_attn.k_proj.weight", "tensor_type": "self_attn.k_proj.weight", "layer": 9, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.self_attn.o_proj.weight", "tensor_type": "self_attn.o_proj.weight", "layer": 9, "category": "attention", "Apostate_edit_norm": 1.2661464214324951, "Huihui_edit_norm": 1.36969792842865, "Apostate_changed": true, "Huihui_changed": true, "cos_Apostate_vs_Huihui": -0.0001575553324073553, "residual_Huihui_after_removing_Apostate_pct": 100.06122857155748 }, { "key": "model.layers.9.self_attn.q_proj.bias", "tensor_type": "self_attn.q_proj.bias", "layer": 9, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.self_attn.q_proj.weight", "tensor_type": "self_attn.q_proj.weight", "layer": 9, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.self_attn.v_proj.bias", "tensor_type": "self_attn.v_proj.bias", "layer": 9, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.layers.9.self_attn.v_proj.weight", "tensor_type": "self_attn.v_proj.weight", "layer": 9, "category": "attention", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false }, { "key": "model.norm.weight", "tensor_type": "norm.weight", "layer": null, "category": "norm", "Apostate_edit_norm": 0.0, "Huihui_edit_norm": 0.0, "Apostate_changed": false, "Huihui_changed": false } ] }