sasa2000/Qwen3-Swallow-8B-RL-v0.2-heretic
Abliterated vs tokyotech-llm/Qwen3-Swallow-8B-RL-v0.2. Looks like abliteration (refusal removal): only the output projections (down_proj, o_proj) changed, in 29 of 36 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.
Tensors changed: Embeddings 0/1, Norms 0/145, MLP 26/108, Attention 29/144, lm_head 0/1
2 derivatives checked
- mradermacher/Qwen3-Swallow-8B-RL-v0.2-heretic-i1-GGUF: Quantization
- mradermacher/Qwen3-Swallow-8B-RL-v0.2-heretic-GGUF: Same weights, converted