Lowkeyss/Glm5.2

Abliterated vs zai-org/GLM-5.2-FP8. Looks like abliteration (refusal removal): only the output projections (o_proj) changed, in 78 of 79 layers. Abliteration edits exactly these matrices to remove a "refusal direction"; everything else is untouched.

Tensors changed: MLP 0/1733, Attention 154/900, lm_head 0/1, Embeddings 0/1, Norms 0/364, other 0/1

2 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse