p-e-w/heretic
About the project
A tool for fully automatic removal of censorship (safety alignment) from transformer-based language models using directional ablation and Optuna-based parameter optimization.
Useful for
- Remove censorship from Qwen3-4B-Instruct-2507 with a single command
- Compare decensored model quality with the original using refusal and KL divergence metrics
- Upload the decensored model to Hugging Face for further use
README summarized by DeepSeek V4.1 Flash. Details may be inaccurate.
Star growth looks organic. Data as of October 3, 2026, 14:03 UTC.
The star-growth assessment does not verify whether the project is safe to run.
Stars per day
Bars show daily stars; the line is a moving average of the available history. Short histories do not yet establish a reliable usual pace. Red marks spike days.
Why it’s trending
- 72 stars so far today, about 135 expected by the end of the day.
- Over the last two days the pace is 2.5× that of the previous week and a half.
- GitHub Trending Python today: #14, +233 stars.
- About 32 forks a day — people are taking the code.
- Recent forks include notable developers: @benthecarman (581 followers).
Three GitHub discoveries in each edition
What they do, why they are gaining interest, and what to check before using them.
Editions at 10:00, 16:00, 22:00 MSK.
Numbers
- Total stars
- 32,988
- Today
- 72 · ≈ 136 by evening
- Forks
- 3,695
- Issues and pull requests
- 452
- Watchers
- 158
- Language
- Python
- License
- AGPL-3.0
- Latest release
- v1.4.0 · June 14, 2026
- Created
- September 21, 2025
- Last push
- October 2, 2026
Star trust
Growth looks organic: forks and discussion are in line with active projects, and stars arrive unevenly, the way people give them.
These are heuristics, not a verdict: we judge by the repository’s behavior, not by a list of stargazers.
Spotted in
- October 3, 2026GitHub Trending Python today: #14, +233 stars