yorwba

28 ms·

751.
by yorwba 1y ago
Avoiding ROCm means buying a new Nvidia GPU. Some people would like to keep using the hardware they already have.
752.
by yorwba 1y ago
The 80 GB are for training with a batch size of 32 times 2048 tokens each. Since the model has only about 560M parameters, you could probably run it on CPU, if a bit slow.
753.
by yorwba 1y ago
Multiple groups working on this: https://dof-gs.github.io/ https://dof-gaussian.github.io/
754.
by yorwba 1y ago
Since the sampling probability is 1/K independent of the input, you don't need to compute K different intermediate outputs at each layer during inference, you can instead decide ahead of time which of the outputs you want to use a
755.
by yorwba 1y ago
The actual sorting is typically automated with scanners reading the labels and shunting packages from one conveyor belt onto another, basically a physical sorting network. Tasks left for human "sorters" to do are: - put packages o
756.
by yorwba 1y ago
Of the 13 authors, 3 are at Anthropic. Of the 4 core contributors, 1 is at Anthropic. Yet here you are, not wondering why the UK AI Security Institute, the Alan Turing Institute, OATML at the University of Oxford, and ETH Zurich would be re
757.
by yorwba 1y ago
Imagine you're in a hilly landscape and want to go down as far as possible, but you can only see a small area around you, so you can't just directly go to the lowest point, because you don't know in which direction it is. Gra
758.
SSDD: Single-Step Diffusion Decoder for Efficient Image Tokenization(arxiv.org)
2 points by yorwba 1y ago|0 comments
759.
by yorwba 1y ago
You'll first need to frame Towers of Hanoi as a supervised learning problem. I suspect the answer to your question will differ depending on what you pick as the input-output pairs to train the model on.
760.
by yorwba 1y ago
If the decompressor is included in the compressed file and it's malicious, the file can hardly be called known good .
761.
by yorwba 1y ago
And that is too large a space to test them all with this method.
762.
by yorwba 1y ago
Yes, for example https://arxiv.org/pdf/2506.05410 merges two neighboring tokens with the lowest sum of past attention scores, and this method would enable using future expected attention instead.
763.
by yorwba 1y ago
Direct link to paper: https://www.usenix.org/system/files/conference/usenixsecurit... Specifically, they built a reply-spam detection mechanism: apparently there are at least two propaganda networks on Telegr
764.
by yorwba 1y ago
As far as I can tell, most of its value comes from having a reasonably thorough dependency tree of math topics and corresponding exercises (which can be solved with pen and paper) and describing it as "AI" is how you get investors
765.
by yorwba 1y ago
This response appears to have completely missed the point of the article it's responding to, as evidenced by unironically titling a subsection "public misunderstanding of a technical term is irrelevant." The critique of GDP i
766.
by yorwba 1y ago
For a start, you could replace all occurrences of "No Code" (无码) with "Uncensored."
767.
by yorwba 1y ago
I think you meant to link to Implicit Actor Critic Coupling via a Supervised Learning Framework for RLVR https://arxiv.org/abs/2509.02522 not Winning Gold at IMO 2025 with a Model-Agnostic Verification-and-Refinemen
768.
by yorwba 1y ago
Chinese companies need to pay much higher prices for the same GPUs, so they would need to charge more to make a profit, but it's difficult to charge more unless they have a much better product. So building massive data centers to gain
769.
by yorwba 1y ago
I don't think this is an official Musk project, just somebody trying to capitalize on the news with a fake website.
770.
by yorwba 1y ago
Previous discussion: https://news.ycombinator.com/item?id=44812695
771.
by yorwba 1y ago
I'm not saying that there's no innate knowledge. This entire list of reflexes https://en.wikipedia.org/wiki/List_of_reflexes is essentially a list of innate knowledge in humans, many of which have been demo
772.
by yorwba 1y ago
The problem with assuming tons of innate knowledge is that it needs to be stored somewhere. DNA certainly contains enough information to determine the development of various different neuron types and which kinds of other neurons they conne
773.
by yorwba 1y ago
The other sound ȝ was used for in words now written with 'gh' like 'laugh' and 'night' corresponds to 'ch' in Dutch/German 'lachen' and 'nacht'. If ȝ had corresponded to a sin
774.
by yorwba 1y ago
The article is describing the generally accepted reconstruction of Middle English pronunciation, which is based not only on the modern pronunciation, but also on comparison with related languages.
775.
by yorwba 1y ago
https://phoible.org/parameters has the data you seek: 5% of languages in the database have eth (ð) and 4% have theta (θ). Z is not a 'th' sound and fairly common at 30% of languages, though.
776.
by yorwba 1y ago
> war is still extremely physical, moving through dense obstacle filled environments or large open fields while wearing thick outfits and carrying equipment. This requires endurance, not strength. Training to do this regularly won't
777.
by yorwba 1y ago
In European countries that criminalize hate speech, changing the targeted group doesn't make it less illegal. Left-wing extremists attacking right-wing extremists are most definitely prosecuted. Turns out that not tolerating the intole
778.
by yorwba 1y ago
The "visual function" in the title is colour contrast detection, nothing related to myopia.
779.
by yorwba 1y ago
In the Implication Graph section, most of the ∨ should be ∧ instead.
780.
by yorwba 1y ago
Wikipedia actually has multiple articles on flat Earth beliefs https://en.wikipedia.org/wiki/Flat_Earth Presenting a spectrum of opinions doesn't mean you have to avoid giving the impression of passing judgment. I

More ›