tvall43/qwen3.5-18b-a3b-reap-coding-heretic-v0

Abliterated vs Flagstone8878/Qwen3.5-18B-REAP-A3B-Coding. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj, out_proj) changed, in 21 of 40 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 0/131, Attention 21/280, MLP 17/2587

3 derivatives checked

On the Hub · Downloads on Model Pulse