EventHorizon-AI/Light-R1-14B-DS-abliterated
Abliterated vs qihoo360/Light-R1-14B-DS. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 47 of 48 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Norms 0/97, Attention 47/336, MLP 47/144, Embeddings 0/1, lm_head 0/1
2 derivatives checked
- mradermacher/Light-R1-14B-DS-abliterated-i1-GGUF: Quantization
- mradermacher/Light-R1-14B-DS-abliterated-GGUF: Quantization