Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Vision language models are blind

vlmsareblind.github.io

Post reply on HN
Print view View original

141–150 of 202 posts

Re: Vision language models are blind

#141

Post by buryat »

joelburget wrote: post #128

Vision Transformers do a shocking amount of compression in the tokenizer. In the [Chameleon paper]( https://arxiv.org/pdf/2405.09818 ) they say the tokenizer "encodes a 512 × 512 image into 1024 discrete tokens from a codebook of size 8192". That's 256 pixels per token (512 * 512 / 1024). If we assume that a pixel is 24 bits (3x 8 bit channels), this implies that they've compressed 256 * 24 = 6144 bits into 13 = (log…

for some reason I started thinking about trying to describe the taste of a fruit to someone who hasn't tried it as something that can be similar to this as a non-visual sensory modal in humans

Re: Vision language models are blind

#142

Post by yantrams »

Tested these problems with llava-v1.6-mistral-7b and the results aren't bad. Maybe I just got lucky with these samples

Intersecting Lines https://replicate.com/p/s24aeawxasrgj0cgkzabtj53rc

Overlapping Circles https://replicate.com/p/0w026pgbgxrgg0cgkzcv11k384

Touching Circles https://replicate.com/p/105se4p2mnrgm0cgkzcvm83tdc

Circled Text https://replicate.com/p/3kdrb26nwdrgj0cgkzerez14wc

Nested Squares https://replicate.com/p/1ycah63hr1rgg0cgkzf99srpxm

Re: Vision language models are blind

#143

Post by simonw »

yantrams wrote: post #142

Tested these problems with llava-v1.6-mistral-7b and the results aren't bad. Maybe I just got lucky with these samples Intersecting Lines https://replicate.com/p/s24aeawxasrgj0cgkzabtj53rc Overlapping Circles https://replicate.com/p/0w026pgbgxrgg0cgkzcv11k384 Touching Circles https://replicate.com/p/105se4p2mnrgm0cgkzcvm83tdc Circled Text https://replicate.com/p/3kdrb26nwdrgj0cgkzerez14wc Nested Squares https://repli…

These are really interesting examples, thanks for sharing.

Re: Vision language models are blind

#144

Post by drodgers »

cs702 wrote: post #7

Wow, that is embarrassingly bad performance for current SOTA models (GPT-4o, Gemini-1.5 Pro, Sonnet-3, Sonnet-3.5), which are advertised and sold as being able to understand images, e.g., for guiding the blind or tutoring children in geometry! The tasks at which they fail are ridiculously simple for human beings, including, for example: * counting the number of times two lines intersect; * detecting whether two circl…

I can't help but read comments like this as:

"My talking dog always makes mistakes on calculus problems: how embarrassingly bad!"

Has the expectation treadmill really advanced so quickly that sub-human performance on any category of problems is now an embarrassment?

Re: Vision language models are blind

#145

Post by ec109685 »

joelburget wrote: post #128

Vision Transformers do a shocking amount of compression in the tokenizer. In the [Chameleon paper]( https://arxiv.org/pdf/2405.09818 ) they say the tokenizer "encodes a 512 × 512 image into 1024 discrete tokens from a codebook of size 8192". That's 256 pixels per token (512 * 512 / 1024). If we assume that a pixel is 24 bits (3x 8 bit channels), this implies that they've compressed 256 * 24 = 6144 bits into 13 = (log…

It’s not as simple as that. If you ask GPT-4o to create a copy of these images, it generally creates one faithfully (e.g. an image with 5 squares will be produced), so it’s “seeing” things reasonably enough.

It doesn’t seem to have the logic though to answer these questions.

The complete data set is here to play around with it yourself: https://huggingface.co/datasets/XAI/vlmsareblind/viewer/defa...

Re: Vision language models are blind

#146

Post by ec109685 »

_vaporwave_ wrote: post #20

It's really interesting that there's a huge performance discrepancy between these SOTA models. In the Olympic logo example, GPT-4o is below the baseline accuracy of 20% (worse than randomly guessing) while Sonnet-3.5 was correct ~76% of the time. Does anyone have any technical insight or intuition as to why this large variation exists?

The question wasn’t “yes or no” but instead required an exact number: https://huggingface.co/datasets/XAI/vlmsareblind/viewer/defa...

Playing around with GPT-4o, it knows enough to make a copy of an image that is reasonable but it still can’t answer the questions.

ChatGPT went down a rabbit hole of trying to write python code, but it took lots of prompting for it to notice its mistake when solving one of the intersecting line questions.

Re: Vision language models are blind

#147

Post by poikroequ »

It's ironic, they fail these seemingly simple tests that are trivial even for a child to solve. Yet, I used Gemini to read a postcard containing handwritten Russian cursive text with lots of visual noise (postmarks and whatnot). It was able to read the text and translate it into English. I didn't even need to tell it the text is Russian.

On the one hand, it's incredible what these LLMs are capable of. On the other hand, they often fall flat on their face with seemingly simple problems like this. We are seeing the same from self driving cars, getting into accidents in scenarios that almost any human driver could have easily avoided.

Re: Vision language models are blind

#148

Post by ClumsyPilot »

godelski wrote: post #87
sweezyjeezy wrote: post #8

Entertaining, but I think the conclusion is way off. > their vision is, at best, like that of a person with myopia seeing fine details as blurry is a crazy thing to write in an abstract. Did they try to probe that hypothesis at all? I could (well actually I can't) share some examples from my job of GPT-4v doing some pretty difficult fine-grained visual tasks that invalidate this. Personally, I rate this paper [1], wh…

> Did they try to probe that hypothesis at all? I think this is a communication issue and you're being a bit myopic in your interpretation. It is clearly an analogy meant for communication and is not an actual hypothesis. Sure, they could have used a better analogy and they could have done other tests, but the paper still counters quite common claims (from researchers) about VLMs. > I could (well actually I can't) sh…

> Second, so what if a test was designed to trick up a model? Shouldn't we be determining when and where models fail? Is that not a critical question in understanding how to use them properly?

People are rushing to build this AI into all kinds of products, and they actively don’t want to know where the problems are.

The real world outside is designed to trip up the model. Strange things happen all the time.

Because software developers have no governing body, no oaths of ethics and no spine someone will end up dead in a ditch from malfunctioning AI.

Re: Vision language models are blind

#149

Post by yantrams »

simonw wrote: post #143
yantrams wrote: post #142

Tested these problems with llava-v1.6-mistral-7b and the results aren't bad. Maybe I just got lucky with these samples Intersecting Lines https://replicate.com/p/s24aeawxasrgj0cgkzabtj53rc Overlapping Circles https://replicate.com/p/0w026pgbgxrgg0cgkzcv11k384 Touching Circles https://replicate.com/p/105se4p2mnrgm0cgkzcvm83tdc Circled Text https://replicate.com/p/3kdrb26nwdrgj0cgkzerez14wc Nested Squares https://repli…

These are really interesting examples, thanks for sharing.

You're welcome. I recently noticed I get better performance with VLMs when the queries are phrased this way - Descriptive Keys instead of explaining the problem in sentences. Similar to COT reasoning that many people claim gives better results, I personally found querying in this sequence - existenceOfEntity, numberOfEntities followed by propertiesOfEntities etc tends to give better results. I haven't verified any of this rigorously so please do take it with a pinch of salt :)

Re: Vision language models are blind

#150

Post by ec109685 »

schneehertz wrote: post #126
pjs_ wrote: post #52

I don't like this paper for the following reasons: - The language is unnecessarily scathing - They repeatedly show data where the models are getting things right 70, 80, 90% of the time, and then show a list of what they call "qualitative samples" (what does "qualitative" mean? "cherry-picked"?) which look very bad. But it got the answer right 70/80/90% of the time! That's hardly "blind"... - Various of the tasks hin…

I am not sure where their experimental data came from. I tested it on GPT-4o using the prompt and images they provided, and the success rate was quite high, with significant differences from the results they provided.

Their examples are here: https://huggingface.co/datasets/XAI/vlmsareblind/viewer/defa...

ChatGPT whiffs completely on very obvious images.

Post reply on HN