adebayoj

6 ms·

1.
by adebayoj 3mo ago
Hey HN! We're Guide Labs and we just launched Clarity, an AI platform powered by our Steerling-8B model. You can: Click any chunk of output the model generates and see the concepts/ideas behind it. Trace the outputs back to the tr
2.
Show HN: Clarity, See what concepts your LLM uses and trace it to training data(guidelabs.ai)
4 points by adebayoj 3mo ago|1 comments
3.
Aligning LLMs at inference time by suppressing internal concepts(guidelabs.ai)
2 points by adebayoj 6mo ago|0 comments
4.
by adebayoj 7mo ago
Hey HN we recently released Steerling-8B, an 8B model designed to be interpretable from the ground up. The model has ~100K concept slots it fills on its own during training, and we can read off what each one means by projecting into vocabul
5.
Extracting 100K concepts from an 8B LLM(guidelabs.ai)
2 points by adebayoj 7mo ago|1 comments
6.
by adebayoj 7mo ago
This part of the claim is involved, so we have future posts to clarify this. And yes, you can remove a prototype and generate again. We show examples in that prism post. In prism, for any token the model generates, you can say, it generated
7.
by adebayoj 7mo ago
It is not hype. You can try the model on huggingface yourself to see its capabilities. My reply here was clarifying that the examples we showed were ones where the model didn't make a mistake. This is intentional, because over the next
8.
by adebayoj 7mo ago
We train the model with `explanations`. Most training asks the model to predict the next token or group of tokens. Our training says, predict the next group of tokens (causal diffusion), but also these tokens should be about {sports/ar
9.
by adebayoj 7mo ago
Take a look at the link in the blogposts. Here is a github link as well: https://github.com/guidelabs/steerling . The model weights are on huggingface, so you can play with it.
10.
by adebayoj 7mo ago
Great questions. We weren't quite explicit about the training data attribution process. We'll discuss this in more detail in future work. We can track down which parts of the training data were interpolated to create that sentence
11.
by adebayoj 7mo ago
You got it exactly right :) And you can update the attribution.md to have it NOT rely on opensource projects that have been compromised. Imagine asking claude code to write a package/function in the style of a codebase that you care ab
12.
by adebayoj 7mo ago
Down to the very exact text chunk in a document! Check this out for an idea of what smaller versions of this style of model can do: https://www.guidelabs.ai/post/prism/ . We'll have more to say soon about it.
13.
by adebayoj 7mo ago
Yes, that is the post that has the most up to date details of the model architecture. Take a look at this: https://github.com/guidelabs/steerling . It has the scaffolding for what you need :)
14.
by adebayoj 7mo ago
You are exactly right, it is guiding the model, during training, with concepts and the dictionary. This is important because dictionary learning for interpretability (post hoc) is not currently reliable: https://www.arxiv.org&#x2
15.
by adebayoj 7mo ago
It is impossible to completely get rid of hallucinations. However, this can tell you exactly why the model hallucinated.
16.
by adebayoj 7mo ago
Exactly! We will have a future post that shows this more granularly over the coming weeks. Here is a post we wrote on how this works at smaller scale: https://www.guidelabs.ai/post/prism/
17.
by adebayoj 7mo ago
Our decomposition allows us to answer question like: for 84 percent of the model's representation, we know it is relying on this concept to give an answer. We can also trace its behavior to the training data that led to it, so that can
18.
by adebayoj 7mo ago
Thanks for pointing this out. LLama 3 8B was trained on ~15T tokens. The Qwen models on 15-18T tokens as well. We trained on 1.35T tokens, and are within shot of these models on benchmarks. We expect to, at the very minimum, match these mod
19.
by adebayoj 7mo ago
Good point. Historically, people have thought that there is a interpretability vs quality/performance tax. This is not true; at least not in this case. Here are a bunch of questions you can answer without any quality degradation with i
20.
by adebayoj 7mo ago
sounds great! Will follow up via email.
21.
by adebayoj 7mo ago
Thanks, it is certainly a first step.
22.
by adebayoj 7mo ago
You are missing a few things, but you got some things right. 1) The is not an SAE in the way you think. It is a combination of a supervised + unsupervised layer that is constrained. An SAE is typically completely unsupervised, and applied p
23.
by adebayoj 7mo ago
It does :) We constrained the model to do exactly this during training: https://www.guidelabs.ai/post/scaling-interpretable-models-8... .
24.
by adebayoj 7mo ago
You got it exactly right. Guilty as charged. Over the coming weeks, we will be showcasing exactly how you can debug all of these examples. I agree that attribution is most useful for debugging and auditing. This is a prime usecase for us. W
25.
by adebayoj 7mo ago
Great questions. We have several posts in the works that will drill down more into these things. The model was actually designed to answer these questions for any sentence (or group of tokens it generates). It can tell you which specific te
26.
by adebayoj 7mo ago
Most interpretability techniques haven't yet to be shown to be useful for everyday model pipelines. However, the field is working hard to change this.
27.
by adebayoj 7mo ago
Completely agree with all your points! Here is what this model does: it `rewrites` the model's activations (during pre-training) into supervised + unsupervised concepts that are then decoded into tokens. So at pre-training, we constrai
28.
by adebayoj 7mo ago
op here. Important point, but I disagree. We see explainability/interpretability as a CORE need for AI safety. We believe you can't align/audit/debug/fix a system that you don't understand. Just to give you som
29.
by adebayoj 7mo ago
op here, I mostly agree with your comment! However, our model does more than this. For any chunk the model generates, it can answer: which concept, in the model's representations, was responsible for that token(s). In fact, we can answ
30.
Show HN: Steerling-8B, a language model that can explain any token it generates(guidelabs.ai)
328 points by adebayoj 7mo ago|91 comments

More ›