megabytes/Jan-v3-4B-base-instruct-heretic
Abliterated vs janhq/Jan-v3-4B-base-instruct. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 28 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: lm_head 0/1, Embeddings 0/1, MLP 25/108, Norms 0/145, Attention 28/144
2 derivatives checked
- mradermacher/Jan-v3-4B-base-instruct-heretic-i1-GGUF: Quantization
- mradermacher/Jan-v3-4B-base-instruct-heretic-GGUF: Same weights, converted