A new open-source tool called Heretic can strip the built-in refusal behavior from locally run large language models with a single command, automating a process that previously demanded manual tuning and deep expertise. Developed by a maintainer using the handle p-e-w and distributed through GitHub, the project applies a technique known as directional ablation—commonly called "abliteration"—to remove the weights associated with a model's tendency to decline certain prompts.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.
Unlocking the full article...
Contact us
Questions, corrections, or ideas? Send us a message here. We aim to reply by email within 24 hours.
Thank you. Your message has been sent. We aim to reply by email within 24 hours.Please enter your email address and a message.We could not send your message. Please try again in a little while.