Peter Robicheaux: evaluation

Transcript context

…Alright, so, yeah, so in that theme one of the big things that we're focusing on is how do we get more out of our pre trained models. And one of the lenses to look at this is through sort of [00:20:00] this, this new requirement for like, how Fine grained visual details and your representations that are extracted from your foundation model. So it's sort of a hook for this Oh, yeah, this is just a list of all the the papers that I'm going to mention I just want to make sure I set an actual paper so you can find it later Yeah, so sort of the big hook here is that I make the claim that LLMs can't see if you go to if you go to Claude or ChatGPT you ask it to see this Watch and tell me what time it is, it fails, right? And so you could say, like, maybe, maybe the Like, this is, like, a very classic test of an LLM, but you could say, Okay, maybe this, this image is, like, too zoomed out, And it just, like, it'll do better if we increase the resolution, And it has easier time finding these fine grained features, Like, where the watch hands are pointing. Nodice. And you can say, okay, well, maybe the model just doesn't know how to tell time from knowing the position of the hands. But if you actually prompt [00:21:00] it textually, it's very easy for it to tell the time. So this to me is proof that these LLMs literally cannot see the position of the watch hands and it can't see those details. So the question is sort of why? And for you anthropic heads out there, cloud fails too. So the, the, my first pick for best paper of 2024 Envision is this MMVP paper, which tries to investigate the Why do LLMs not have the ability to see fine grained details? And so, for instance, it comes up with a lot of images like this, where you ask it a question that seems very visually apparent to us, like, which way is the school bus facing? And it gets it wrong, and then, of course, it makes up details to support its wrong claim. And so, the process by which it finds these images is sort of contained in its hypothesis for why it can't. See these details. So it hypothesizes that models that have been initialized with, with Clip as their vision encoder, they don't have fine grained details and the, the features extracted using Clip because Clip sort of doesn't need to find these fine grained [00:22:00] details to do its job correctly, which is just to match captions and images, right? And sort of at a high level, even if ChatGPT wasn't initialized with Clip and wasn't trained contrastively at all. The vision encoder wasn't trained contrastively at all. Still, in order to do its job of capturing the image it could do a pretty good job without actually finding the exact position of all the objects and visual features in the image, right? ively at all. Still, in order to do its job of capturing the image it could do a pretty good job without actually finding the exact position of all the objects and visual features in the image, right? So This paper finds a set of difficult images for these types of models. And the way it does it is it looks for embeddings that are similar in clip space, but far in DynaV2 space. So DynaV2 is a foundation model that was trained self supervised purely on image data. And it kind of uses like some complex student teacher framework, but essentially, and like, it patches out like certain areas of the image or like crops with certain areas of the image and tries to make sure that those have consistent representations, which is a way for it to learn very fine grained visual features. And so if you take things that are very close in clip space and very far in DynaV2 space, you get a set of images [00:23:00] that Basically, pairs of images that are hard for a chat GPT and other big language models to distinguish. So, if you then ask it questions about this image, well, as you can see from this chart, it's going to answer the same way for both images, right? Because to, to, from the perspective of the vision encoder, they're the same image. And so if you ask a question like, how many eyes does this animal have? It answers the same for both. And like all these other models, including Lava do the same thing, right? And so this is the benchmark that they create, which is like finding clip, like clip line pairs, which is pairs of images that are similar in clip space and creating a data set of multiple choice questions based off of those. And so how do these models do? Well, really bad. Lava, I think, So, so, chat2BT and Jim and I do a little bit better than random guessing, but, like, half of the performance of humans who find these problems to be very easy. Lava is, interestingly, extremely negatively correlated with this dataset. It does much, much, much, much worse [00:24:00] than random guessing, which means that this process has done a very good job of identifying hard images for, for Lava, specifically. And that's because Lava is basically not trained for very long and is initialized from Clip, and so You would expect it to do poorly on this dataset. So, one of the proposed solutions that this paper attempts is by basically saying, Okay, well if clip features aren't enough, What if we train the visual encoder of the language model also on dyno features? And so it, it proposes two different ways of doing this. One, additively which is basically interpolating between the two features, and then one is interleaving, which is just kind of like training one on the combination of both features. So there's this really interesting trend when you do the additive mixture of features. tures, and then one is interleaving, which is just kind of like training one on the combination of both features. So there's this really interesting trend when you do the additive mixture of features. So zero is all clip features and one is all DynaV2 features. So. It, as you, so I think it's helpful to look at the right most chart first, which is as you increase the number of DynaV2 features, your model does worse and worse and [00:25:00] worse on the actual language modeling task. And that's because DynaV2 features were trained completely from a self supervised manner and completely in image space. It knows nothing about text. These features aren't really compatible with these text models. And so you can train an adapter all you want, but it seems that it's in such an alien language that it's like a very hard optimization for this. These models to solve. And so that kind of supports what's happening on the left, which is that, yeah, it gets better at answering these questions if as you include more dyna V two features up to a point, but then you, when you oversaturate, it completely loses its ability to like. Answer language and do language tasks. So you can also see with the interleaving, like they essentially double the number of tokens that are going into these models and just train on both, and it still doesn't really solve the MMVP task. It gets Lava 1. 5 above random guessing by a little bit, but it's still not close to ChachiPT or, you know, Any like human performance, obviously. [00:26:00] So clearly this proposed solution of just using DynaV2 features directly, isn't going to work. And basically what that means is that as a as a vision foundation model, DynaV2 is going to be insufficient for language tasks, right? So my next pick for best paper of 2024 would be Florence 2, which tries to solve this problem by incorporating not only This dimension of spatial hierarchy, which is to say pixel level understanding, but also in making sure to include what they call semantic granularity, which ends up, the goal is basically to have features that are sufficient for finding objects in the image, so they're, they're, they have enough pixel information, but also can be talked about and can be reasoned about. And that's on the semantic granularity axis. So here's an example of basically three different paradigms of labeling that they do. So they, they create a big dataset. One is text, which is just captioning. And you would expect a model that's trained [00:27:00] only on captioning to have similar performance like chat2BT and like not have spatial hierarchy, not have features that are meaningful at the pixel level.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.