497
submitted 4 days ago* (last edited 4 days ago) by [email protected] to c/[email protected]

In large language model (LLM) pretraining, data quality is believed to determine model quality. In this paper, we re-examine the notion of "quality" from the perspective of pre- and post-training co-design. Specifically, we explore the possibility that pre-training on more toxic data can lead to better control in post-training, ultimately decreasing a model's output toxicity. First, we use a toy experiment to study how data composition affects the geometry of features in the representation space. Next, through controlled experiments with Olmo-1B models trained on varying ratios of clean and toxic data, we find that the concept of toxicity enjoys a less entangled linear representation as the proportion of toxic data increases. Furthermore, we show that although toxic data increases the generational toxicity of the base model, it also makes the toxicity easier to remove. Evaluations on Toxigen and Real Toxicity Prompts demonstrate that models trained on toxic data achieve a better trade-off between reducing generational toxicity and preserving general capabilities when detoxifying techniques such as inference-time intervention (ITI) are applied. Our findings suggest that, with post-training taken into account, bad data may lead to good models.

(page 2) 50 comments
sorted by: hot top new old
[-] [email protected] 8 points 4 days ago

My hope was that AI would, at least, bear some disgust for the worst of humanity. My new fear is that AI will bear disgust for humanity.

[-] [email protected] 8 points 4 days ago

Not to anthropomorphize LLMs, but.... Like a vaccine?

[-] [email protected] 3 points 4 days ago

Kinda of actually

[-] [email protected] 7 points 4 days ago* (last edited 4 days ago)

It's like how vaccinations protect us from illnesses.

[-] [email protected] 7 points 4 days ago

Interesting training strategy. Makes a lot of sense intuitively. Worried this makes the model even more susceptible to prompt injections. Feels like this method adds more attack vectors? It's unfortunate they didn't attempt to test the long term hardness and stability, though it's probably beyond their scope.

load more comments (2 replies)
[-] [email protected] 2 points 3 days ago

Based and hopepilled

[-] [email protected] 1 points 3 days ago

This is not surprising if you've studied anything on machine learning or even just basic statistics. Consider if you are trying to find out the optimal amount of a thickener to add to a paint formulation to get it to flow the amount you want. If you add it at 5%, then 5.1%, then 5.2%, it will he hard to see how much of the difference between those batches is due to randomness or measurement uncertainty than if you see what it does at 0%, then 25% then 50%. This is a principle called Design of Experiments (DoE) in traditional statistics, and a similar effect happens when you are training machine learning models- datapoints far outside the norm increase the ability of the model to predict within the entire model space (there is some nuance here, because they can become over-represented if care isn't taken). In this case, 4chan shows the edges of the English language and human psychology, like adding 0% or 50% of the paint additives rather than staying around 5%.

At least that's my theory. I haven't read the paper but plan to read it tonight when I have time. At first glance I'm not surprised. When I've worked with industrial ML applications, processes that have a lot of problems produce better training data than well controlled processes, and I have read papers on this subject where people have improved performance of their models by introducing (controlled) randomness into their control setpoints to get more training data outside of the tight control regime.

[-] [email protected] 1 points 3 days ago* (last edited 3 days ago)

I say it's simply easier to recognize something when you've seen more examples of it.

If you're training an image discriminator on apples, bananas, oranges, pears and penises, it will inevitably do better overall if 10-30% of the images it trains on are penises, rather than 0.01% penises - even if in operation it is only expected to encounter dick pics very rarely.

[-] [email protected] 3 points 4 days ago

4chan is fun!

load more comments
view more: ‹ prev next ›
this post was submitted on 09 Jun 2025
497 points (96.8% liked)

Technology

71355 readers
3186 users here now

This is a most excellent place for technology news and articles.


Our Rules


  1. Follow the lemmy.world rules.
  2. Only tech related news or articles.
  3. Be excellent to each other!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
  9. Check for duplicates before posting, duplicates may be removed
  10. Accounts 7 days and younger will have their posts automatically removed.

Approved Bots


founded 2 years ago
MODERATORS