prithivMLmods/Megatron-Opus-14B-2.0
Abliterated vs microsoft/phi-4. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 40 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/81, MLP 40/40, Attention 40/40, lm_head 0/1
5 derivatives checked
- mradermacher/Megatron-Opus-14B-2.0-i1-GGUF: Quantization
- mradermacher/Megatron-Opus-14B-2.0-GGUF: Quantization
- DevQuasar-2/prithivMLmods.Megatron-Opus-14B-2.0-GGUF: Same weights, converted
- prithivMLmods/Megatron-Opus-14B-2.0-GGUF: Quantization
- Xiaojian9992024/Phi-4-Megatron-Empathetic: LoRA-style