p-e-w/heretic
View on GitHub ↗ ★ 37,748 Python
Training-free language model modification: Remove censorship, suppress slop, reduce sycophancy, and more
26 mentions on Hacker News by 19 people: 2 stories, 24 comments · first 2025-11-16 · latest 2026-11-01
20082026
-
…simply moving the location they upload stuff to. To use an old analogy, it's not rocket science: https://github.com/p-e-w/heretic
comment by walrus01 · 2026-11-01 · in Memdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard -
…If someone breaks your window, then they are responsible, not the hammer they use to do so. [1]: https://github.com/p-e-w/heretic
comment by Retro_Dev · 2026-09-22 · in Claude Opus 5.5 -
https://github.com/p-e-w/heretic/blob/3521f8648a0dccf6e12a92...
comment by b5n · 2026-09-21 · in Heretic removes restrictions from language models -
Here's a widely used toolkit: https://github.com/p-e-w/heretic Distributing the vectors themselves isn't (yet) common practice, because people have gotten used to just putting the full modified weights up on HuggingFace's huge free storage.
comment by wren6991 · 2026-09-20 · in Pirate Face Rescues LLM Models from Deletion -
…of undoing the "for your own safety" lobotomy than explicitly specializing in unsafe things. [1] https://github.com/p-e-w/heretic
comment by snackbroken · 2026-09-08 · in We have a year to fix security everywhere -
…LLMs. They have some niche non-coding uses for things that a commercial LLM will never talk about. https://github.com/p-e-w/heretic
comment by walrus01 · 2026-08-30 · in Hy4 preview -
…get you thrown into prison if you published them in Mandarin on the domestic Chinese internet. https://github.com/p-e-w/heretic As a side note on this, if you see the reference in the screenshot in the link above to the harmful behaviors…
comment by walrus01 · 2026-07-27 · in Kimi-K3 on HuggingFace -
You can find the current state-of-art tool for censorship removal here: https://github.com/p-e-w/heretic
comment by kouteiheika · 2026-07-25 · in UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities -
…not, because it's pretty trivial to uncensor[1] any open-weight model and make it not refuse. [1]: https://github.com/p-e-w/heretic
comment by kouteiheika · 2026-07-22 · in Laguna S 2.1 -
Once we've got the weights, anything is possible. https://github.com/p-e-w/heretic
comment by matheusmoreira · 2026-07-21 · in OpenAI and Hugging Face address security incident during model evaluation -
https://github.com/p-e-w/heretic
comment by ls612 · 2026-06-30 · in Claude Sonnet 5 -
They are open weight, so you can abliterate: https://github.com/p-e-w/heretic You can finetune and mould it to whatever you want.
comment by dannyw · 2026-06-14 · in GLM 5.2 Is Out -
…weekend side project with an off-the-shelf tool[1]. (Try it. It's fun. I've done it myself.) [1]: https://github.com/p-e-w/heretic
comment by kouteiheika · 2026-06-12 · in Kimi K2.7-Code: open-source coding model with better token efficiency -
Sounds like something that heretic or similar might be useful for? https://github.com/p-e-w/heretic
comment by justinclift · 2026-06-12 · in Kimi K2.7-Code: open-source coding model with better token efficiency -
heretic maintainer: https://github.com/p-e-w/heretic the fun bits are in another branch or PRs
comment by electroglyph · 2026-05-16 · in DeepSeek-V4-Flash means LLM steering is interesting again -
…is this?" axis internally when refusing queries that can be identified and interfered with. [0] https://github.com/p-e-w/heretic
comment by roenxi · 2026-05-08 · in Teaching Claude Why -
…problem. If you wait a few days after a new model release, someone will have made a heretic ( https://github.com/p-e-w/heretic ) version with the censorship removed, so in a way the only use for censorship now is to avoid lawsuits, not…
comment by hleszek · 2026-05-02 · in Refusal in Language Models Is Mediated by a Single Direction -
…who makes sure that you really can access the full model capabilities. Censorship is the devil. https://github.com/p-e-w/heretic
comment by Der_Einzige · 2026-05-02 · in Refusal in Language Models Is Mediated by a Single Direction -
…to answer you can uncensor it (and this is getting easier and more automated day-by-day[1]). [1] -- https://github.com/p-e-w/heretic
comment by kouteiheika · 2026-05-02 · in DeepSeek V4 – almost on the frontier -
Wonder if stuff like this would affect it? https://github.com/p-e-w/heretic Guessing it probably would?
comment by justinclift · 2026-04-20 · in Kimi K2.6: Advancing open-source coding -
You may be interested in heretic. People often post models to hf that have been un-censored https://github.com/p-e-w/heretic
comment by boredatoms · 2026-04-16 · in Qwen3.6-35B-A3B: Agentic coding power, now open to all -
…tips yet?) I don't normally do the whole "abliterated" thing (dealignment) but after discovering https://github.com/p-e-w/heretic , I was too tempted to try it with this model a couple days ago (made a repo to make it easier, actually)…
comment by pmarreck · 2026-04-05 · in Gemma 4 on iPhone -
…comment, what this project wants to achieve has been done for a while and it's called Heretic: https://github.com/p-e-w/heretic (Not vibecode by a twitter influgrifter)
comment by littlestymaar · 2026-03-06 · in A tool that removes censorship from open-weight LLMs