Qwen3.6 27B: HauhauCS Aggressive, Recovered Safetensors 💬 Community: Join the Abliterlitics Discord for discussion, model releases and support. Forensic analysis by Abliterlitics, open source abliteration forensics toolkit Recovered HuggingFace safetensors from the Q8 K P quantised GGUF published by HauhauCS using ungguf, our GGUF to safetensors conversion tool. I ran the full forensic suite: benchmarks, safety evaluation, KL divergence, and weight analysis. Then I compared against four other abliteration techniques applied to the same base model: Heretic, Huihui, AEON, and Abliterix. HauhauCS used an abliteration tool called "Reaper Abliteration," which was shown to be plagiarised from Heretic under AGPL 3.0 with all attribution stripped and relicensed to PolyForm Noncommercial. Based on our analysis of the recovered source code, on top of the Heretic derived core, Reaper adds subspace level rank k ablation, per component continuous ablation curves, SOM clustering for multi directional refusal discovery, and several other techniques. The model was exported as Q8 K P GGUF, which we converted back to safetensors with ungguf. The weights therefore carry two layers of modification. R…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy