HackerTrans
TopNewTrendsCommentsPastAskShowJobs

E-Reverance

657 karmajoined 6 lat temu

Submissions

A Theory of Contrastive Learning with Natural Images

arxiv.org
4 points·by E-Reverance·przedwczoraj·0 comments

Lingbot Vision

technology.robbyant.com
2 points·by E-Reverance·3 dni temu·0 comments

Zero-Flow Encoders

arxiv.org
1 points·by E-Reverance·4 dni temu·0 comments

Dispersion loss counteracts embedding condensation in small language models

chenliu-1996.github.io
42 points·by E-Reverance·7 dni temu·8 comments

Meshtryoshka: Differentiable Mesh Rendering for Unbounded Scenes

danielxu9393.github.io
2 points·by E-Reverance·11 dni temu·0 comments

Simplified Sparse Attention via Gist Tokens

arxiv.org
4 points·by E-Reverance·11 dni temu·0 comments

Continuity-Enhancing Degree Elevation and Splits

graphics.cs.utah.edu
2 points·by E-Reverance·13 dni temu·0 comments

Forward Self Models

jagilley.github.io
2 points·by E-Reverance·14 dni temu·0 comments

Tapered Language Models

arxiv.org
3 points·by E-Reverance·18 dni temu·0 comments

Everything is logarithms

alexkritchevsky.com
316 points·by E-Reverance·19 dni temu·83 comments

Explaining Attention with Program Synthesis

arxiv.org
2 points·by E-Reverance·23 dni temu·0 comments

Sparsely gated tiny linear experts

arxiv.org
3 points·by E-Reverance·w zeszłym miesiącu·0 comments

Training-Free Single-Image Diffusion Models

haojunqiu.github.io
6 points·by E-Reverance·w zeszłym miesiącu·0 comments

Simple input–output dependencies explain neuronal activity

nature.com
4 points·by E-Reverance·w zeszłym miesiącu·0 comments

Self Teaching Autoencoder

the-puzzler.github.io
2 points·by E-Reverance·2 miesiące temu·0 comments

Distance Marching for Generative Modeling

arxiv.org
1 points·by E-Reverance·2 miesiące temu·0 comments

A Theory of Generalization in Deep Learning

arxiv.org
4 points·by E-Reverance·2 miesiące temu·0 comments

Neural similarity predicts whether strangers become friends

nature.com
4 points·by E-Reverance·2 miesiące temu·0 comments

Generation Is Required for Data-Efficient Perception

arxiv.org
1 points·by E-Reverance·2 miesiące temu·0 comments

Flow Map Learning via Nongradient Vector Flow [pdf]

openreview.net
26 points·by E-Reverance·3 miesiące temu·2 comments

comments

E-Reverance
·23 dni temu·discuss
I don't think the issue is that they used ai behind the scenes, but there is an implicit proof of work from forcing it beyond the style you'd expect. I for one roll my eyes whenever I see that specific kind of rounded corner, frosted glass ui and layout choices. It looks like someone trying to superficially/ham-fistedly trying to replicate "good taste" without actually having a good model of taste, its quite uncanny/bootleg.
E-Reverance
·28 dni temu·discuss
"At the same time, China is also the world's leading producer of electric cars..."

Kind of interesting for a professionally branded company to use "..." like that
E-Reverance
·w zeszłym miesiącu·discuss
I think its worth emphasizing that his argument isn't completely against generative ai, but rather its environment. Although I don't see why it would be impossible for something like an LLM to learn some sort of self-play within its context window
E-Reverance
·w zeszłym miesiącu·discuss
I don't completely disagree but its worth noting how new a lot of the empirical evidence in favour of LLMs are, so its not impossible to be a tad ignorant of the present
E-Reverance
·w zeszłym miesiącu·discuss
No

> It uses 384 routed experts (top-8) with hybrid attention (full-attention + sliding-window 128 at 6:1 ratio) over 70 layers (1 dense + 69 MoE)

https://recipes.vllm.ai/XiaomiMiMo/MiMo-V2.5-Pro
E-Reverance
·2 miesiące temu·discuss
> P.S.: I like discussing such topics. If anyone knows a forum or discord with like-minded people, please let me know :)

Unironically twitter (and only use the "Following" tab as opposed to the "For You")

Make an account that only follows university affiliated researchers with less than 1000 followers. In my experience discord servers get suffocated by beginners and crackpots because conversations don't naturally self-organize into their own threads.
E-Reverance
·2 miesiące temu·discuss
> But I am a bit reasurred that at least my job won't be fully replaced with AI :)

I honestly can't comment with certainty that training from videos alone and whatever tokenization scheme they're using will ever get perfect dynamics.

However it is worth noting that transformers can do a pretty good job at learning dynamics with the right pipeline (not video): https://arxiv.org/pdf/2605.15305 https://arxiv.org/pdf/2605.09196

My point here being that representationally, it might be possible to learn good dynamics without a radically different approach/arch. There are already models that extract 3D tracking points from videos, so they could possibly be leveraged for learning dynamics (which on its own gives precedent for end-to-end approaches also possibly working).
E-Reverance
·2 miesiące temu·discuss
FYI it’s not an Arab country
E-Reverance
·2 miesiące temu·discuss
Interesting comment from him:

"

SPIEGEL: So you don't consider Collins to be a true scientist?

Venter: Let's just say he's a government administrator.

"

https://www.science.org/content/blog-post/craig-venter-venti...
E-Reverance
·2 miesiące temu·discuss
They factorize the distribution in which they are trained on which is essentially generalization

https://arxiv.org/abs/2602.02385
E-Reverance
·3 miesiące temu·discuss
Yeah I wouldn't be surprised if journalist are getting high on their own supply of resentment and fear mongering
E-Reverance
·3 miesiące temu·discuss
People not wanting their jobs be automated is different from not yearning for automation as a principle. Most people want or (at least don't mind) elevators, tap water, dishwashers, traffic lights, electrical fuses, sliding doors, etc. Its a very general term
E-Reverance
·3 miesiące temu·discuss
It does?

" Software brain is powerful stuff. It’s a way of thinking that basically created our modern world. Marc Andreessen, the literal embodiment of software brain, called it in 2011 when he wrote the piece “Why software is eating the world” as an op-ed in The Wall Street Journal. But software thinking has been turbocharged by AI in a way that I think helps explain the enormous gap between how excited the tech industry is about the technology and how regular people are growing to dislike it more and more over time. "
E-Reverance
·3 miesiące temu·discuss
Maybe a nitpicky HN comment, but why are we lumping the term automation with very recent grievances about certain kinds of automation
E-Reverance
·3 miesiące temu·discuss
Try refreshing or something, still works for me
E-Reverance
·3 miesiące temu·discuss
I think its endearing
E-Reverance
·3 miesiące temu·discuss
I might be misinterpreting but the LUAR model (which is a transformer) seems to do decently well

https://www.nature.com/articles/s41599-025-06340-3/figures/2
E-Reverance
·3 miesiące temu·discuss
https://www.bbc.com/news/articles/c0mgndkklvmo
E-Reverance
·3 miesiące temu·discuss
> 75%

Wouldn’t an absolute number make more sense to show than a percent. 75% is pretty good in some places in the world (not justifying the discrepancy though)
E-Reverance
·3 miesiące temu·discuss
Could this be used for batch filtering?