Hemmingway-1-heretic-GGUF

Hemmingway-1-heretic-GGUF

This repository contains static GGUF quantizations of Hemmingway-1-heretic, created by applying the arelath/Hemmingway-1-heretic-adapter to Altworld/Hemmingway-1 (a 27B model built on Qwen3.8-27B) and quantizing for local inference with llama.cpp.

Hemmingway-1 is tuned specifically to write like a real person—skipping preambles, lecture notes, and corporate filler. This heretic version decensors the weights via targeted representation abliteration using Heretic v2.0.0, reducing refusals to near-zero while preserving high-EQ conversational voice.

---

Quantization Variants

---

Decensoring & Abliteration Details

Created using Heretic v2.0.0.dev0 against Altworld/Hemmingway-1.

Abliteration Parameters

• direction_index: 38.05
• attn.o_proj.max_weight: 1.03 (position: 41.01, min_weight: 1.02, min_distance: 28.56)
• mlp.down_proj.max_weight: 1.41 (position: 51.07, min_weight: 1.39, min_distance: 35.29)

Refusal Comparison

---

Quickstart & Usage

1. llama.cpp CLI

# Clone and build llama.cpp, then run:
./llama-cli \
  -m Hemmingway-1-heretic-Q4_K_M.gguf \
  -p "<|im_start|>system\nYou are a direct, empathetic writing assistant.<|im_end|>\n<|im_start|>user\nWrite the text I send my landlord about the broken boiler.<|im_end|>\n<|im_start|>assistant\n" \
  -n 512 \
  -ngl 99 \
  -c 8192

Credits & Licensing

Base Weights: Altworld/Hemmingway-1 by Altworld, built on Qwen/Qwen3.8-27B.

Abliteration Adapter: arelath/Hemmingway-1-heretic-adapter by arelath.

License: Governed by the original model's license terms: CC BY-NC 4.0 (Free for personal and non-commercial research use; commercial usage requires agreement with Altworld).