HF modelNEW

OS-Software / llm-jp-4.1-32b-a3b-thinking-uncensored-heretic-GGUF

A decensored, safety‑reduced version of llm-jp/llm‑jp‑4.1‑32b‑a3b‑thinking using custom Heretic Abliteration parameters (target layers 11, 13–18, 20–23 with varied layer weights, steer_bad_behavior_weight 0.03, etc.) that matches the original model’s MMLU and HumanEval scores while noting higher risk of harmful content and intended only for research or safety testing.

Description in other languages

Statistics

Statistics as of Nov 16, 2026

Hugging Face downloads (last 30 days)
6,000
Hugging Face likes
1
Listed
Nov 3, 2026

Popular in the same category

Resources in similar research fields

Related slides and articles

Ones that name this resource in their titles, and ones close to its description

Learn from slides & articles