/AI17h ago

Obliteratus Ablates LLM Refusal Weights With One-Click SVD Analysis

94936461923K

Removes LLM censorship with a single click. credit via @elder_plinius bro,

Obliteratus, the open-source weapon that hunts down the exact weights causing refusals and surgically projects them out, is helpful for cybersec.

Obliteratus uses advanced techniques (SVD, mean-difference analysis) to detect refusal behaviors and remove them at the weight level without retraining One-click ablation, Gradio playground for no-code use, Reversible steering vectors,

Supports 100+ Hugging Face models, Multi-GPU ready.

- http://github.com/elder-plinius/OBLITERATUS

9:48 AM ยท Jun 7, 2026 ยท 23K Views
Sentiment

Users thank the creators of the Obliteratus open-source tool for ablating LLM refusal weights because the world needs uncensored AI free from narrative control by certain groups.

Pos
100.0%
Neg
0.0%
1 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS28

@0x0SojalSec @elder_plinius Thank you for sharing.

World needs uncensored Ai.

Certain groups are controlling the narrative.

8hViews 28

@0x0SojalSec @elder_plinius ablation is such a precise word ๐ŸคŒ๐Ÿผ

8hViews 12