prithivMLmods/Megatron-Opus-14B-2.1
Abliterated vs microsoft/phi-4. Looks like abliteration (refusal removal): only the output projections (down_proj, embed_tokens, 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 1/1, Norms 0/81, MLP 40/40, Attention 40/40, lm_head 1/1
3 derivatives checked
- mradermacher/Megatron-Opus-14B-2.1-i1-GGUF: Quantization
- DevQuasar-9/prithivMLmods.Megatron-Opus-14B-2.1-GGUF: Same weights, converted
- mradermacher/Megatron-Opus-14B-2.1-GGUF: Quantization