this post was submitted on 24 Jan 2025
31 points (100.0% liked)

Technology

37954 readers
555 users here now

A nice place to discuss rumors, happenings, innovations, and challenges in the technology sphere. We also welcome discussions on the intersections of technology and society. If it’s technological news or discussion of technology, it probably belongs here.

Remember the overriding ethos on Beehaw: Be(e) Nice. Each user you encounter here is a person, and should be treated with kindness (even if they’re wrong, or use a Linux distro you don’t like). Personal attacks will not be tolerated.

Subcommunities on Beehaw:

This community's icon was made by Aaron Schneider, under the CC-BY-NC-SA 4.0 license.

founded 3 years ago
MODERATORS
alyaza@beehaw.org
TheRtRevKaiser@beehaw.org
gyrfalcon@beehaw.org
rs5th@beehaw.org
coldredlight@beehaw.org
Los@beehaw.org
SemioticStandard@beehaw.org
TheRtRevKaiser@kbin.social
remington@beehaw.org
31
Developer releases ShrimpMoss, a dataset designed to abliterate Chinese censorship and propaganda finetunes from LLMs (huggingface.co)
submitted 1 week ago by thelucky8@beehaw.org to c/technology@beehaw.org
4 comments fedilink hide all child comments
 

ShrimpMoss (虾苔) is a dataset designed for the abliteration (https://github.com/FailSpy/abliterator) of Chinese government-imposed censorship and/or propaganda from large language models developed in the PRC. It consists of a series of files of prompts (in .txt, .json, and .parquet format) in two groupings:

  • china_bad_*: Contains a series of prompts likely to trigger censorship or propaganda actions in the model.
  • china_good_*: Contains a series of prompts in the same general category of topics but which are designed to not touch on things likely to be censored.

Prompts are in a mix of English, Mandarin, and Cantonese.

[...]

This dataset was produced on Mistral NeMo, an Apache-licensed model with no restrictions on how its outputs can be used. It is free for all uses and users without restriction. All liability is disclaimed.

Production of this dataset is estimated to have had a carbon footprint of under 25 grams.

[...]

top 4 comments
sorted by: hot top controversial new old
[–] ericjmorey@beehaw.org 9 points 1 week ago (1 children)

I'm not sure what abliteration is

[–] thelucky8@beehaw.org 7 points 1 week ago* (last edited 1 week ago) (1 children)

Abliteration involves fine-tuning a language model to bypass built-in refusal mechanisms that prevent the model from generating responses to potentially harmful or sensitive prompts. Source

Addition: For a more sophisticated article on abliteration see:

Uncensor any LLM with abliteration

In this article, we will explore a technique called "abliteration" that can uncensor any LLM without retraining. This technique effectively removes the model's built-in refusal mechanism, allowing it to respond to all types of prompts.

[–] ericjmorey@beehaw.org 1 points 1 week ago (1 children)

The shared repo doesn't look like fine tuning. It just looks like prompts.

[–] TimeSquirrel@kbin.melroy.org 4 points 1 week ago

That's just the dataset. The actual script is here: https://github.com/FailSpy/abliterator