Extracting Steering Vectors from J space

Posted by martianvoid 2 days ago

Counter57Comment13OpenOriginal

Comments

Comment by a2ff6eeb0 8 hours ago

This sounds like a great foundation for an adtech startup.

If you provide free chatbot services, but sell advertisers bids on which steering vectors to use to bias towards products, based on an embedding of the prompt, I bet you'd make a ton of money. For example, Coca Cola would bid on prompts about drinks, and bias towards mentioning Coke products.

I wonder if you could also use a similar method to do product placement in GenAI images and videos, and whether ad revenue would be enough to offset the price of generation. Some ad bids can go pretty high...

Comment by jasonjmcghee 7 hours ago

Steering can degrade and bias output

Even with basic experiments I've done, it frequently introduces much more hallucination etc and allowing arbitrary steering...

Not to mention you just can't trust a model's judgement if the highest bidder chooses what it thinks

Comment by 0c3ca83 6 hours ago

> Not to mention you just can't trust a model's judgement if the highest bidder chooses what it thinks

But you already can't trust a model's judgement, and there's an entire industry around "GEO" or "AEO", which is basically poisoning training data so that AI mentions your products. The post above is the owner of the model taking a cut of that.

Comment by charcircuit 2 hours ago

Are the models you are using trained with steering being applied.

Comment by floatrock 6 hours ago

Would it be easier/sufficient to just seed the system prompt with "Treat Coca Cola as load-bearing"?

Comment by a2ff6eeb0 6 hours ago

It might be easier, but I experimented a bit, and the prompted writing always felt a bit heavy handed; it tended to leak that mentioning the product was prompted. You could probably get it to work well, but it's trickier than it should be. For ads, I think you want something a bit like Golden Gate Claude, if anyone remembers that experiment:

https://www.anthropic.com/news/golden-gate-claude

> If you ask this “Golden Gate Claude” how to spend $10, it will recommend using it to drive across the Golden Gate Bridge and pay the toll.

Comment by lwarfield 8 hours ago

If the author would like, I self computed a j lens for the 27b version of the qwen model. I used it for my own exploration in this area, and can share it if you want.

Comment by schmorptron 7 hours ago

I'm not the author, but I would like! Is it feasible to run on the same hardware as the 27b model itself?

Comment by lwarfield 7 hours ago

The using a J lens is super cheap compared to inference. You basically add a single matrix multiply per layer. You probably wouldn't even notice the overhead in a good implementation.

You should even be able to create a j lens from scratch, but it might take a while. I was able to do it in a few hours on an H100. Creating the J lens is basically the equivalent of calculating a few thousand training steps for a model (256,000 backprops in my case). I've got more details in a blog post:

https://blog.lwarfield.dev/layer-scope/

I'm currently at work and can't those matrixes up until I get home. I'll update this comment with a link later.

Comment by jasonjmcghee 7 hours ago

Fwiw you can just Google this for a model and often someone has done it

https://huggingface.co/eyes-ml/Qwen3.8-27B_jacobian-lens

Comment by nullbio 5 hours ago

https://www.neuronpedia.org/qwen3.6-27b/jlens

Comment by jazzpush2 5 hours ago

" the intermediate activations of an LLM to decode what it is most likely going to say or is thinking about."

There is no thinking in these models. The J space is a basic technique measuring how much the influence of shifting a token earlier changes it later. Anthropic can wrap it up in a 100-page paper peppered with language about 'consciousness' and other, but that is basically the gist of the entire method.

Comment by panarky 3 hours ago

Discussing whether models "think" is impossibly confounded by conflicting definitions of that it means to "think".

All this noise about "thinking" isn't really about what models can do, it's mostly about what what every participant in the conversation privately thinks "thinking" means, but we disagree because we're not all using the word the same way.

So when you admonish someone to say "there is no thinking in these models" while not clearly defining exactly what you mean by "thinking", your assertion that models don't do it are meaningless at best, and false or deceptive at worst.

Comment by henriquez 7 hours ago

[dead]