martin palazzo @boardsofdata

science & startup worker. high dimensional space Joined November 2011

Tweets

6K
Followers

1K
Following

2K
Likes

19K

Simone Scardapane @s_scardapane

2 weeks ago

*The Origins of Representation Manifolds in LLMs* by Modell et al. They study the presence of "interpretable features" in LLMs embedded as manifolds and how their geometry connects to the internal representations of the models. arxiv.org/abs/2505.18235

15 162 1K 52K 689

Download Image

alphaXiv @askalphaxiv

2 weeks ago

This new research showed empirically that KL divergence is the most accurate predictor of catastrophic forgetting in LLMs! And since forgetting tracks KL drift from the base, methods that keep KL small tend to forget less too.

7 54 519 67K 385

Download Image

j⧉nus @repligate

2 weeks ago

HOW INFORMATION FLOWS THROUGH TRANSFORMERS Because I've looked at those "transformers explained" pages and they really suck at explaining. There are two distinct information highways in the transformer architecture: - The residual stream (black arrows): Flows vertically through…

j⧉nus @repligate

3 weeks ago

7 11 159 214K 97

45 296 2K 263K 3K

Download Image

Carlos Sarraute ⚡️ @ch4rleston

3 weeks ago

💡 Me encantaron las sesiones de pósters en el Simposio Científico de IA y Aplicaciones #SCIAA2025! Había muchos trabajos interesantes. En particular quiero destacar dos proyectos que me llamaron la atención: 🔹 Impacto del Prompt Engineering en el rendimiento de LLMs: estudio…

2 4 10 2K 1

Download Image

martin palazzo @boardsofdata

3 weeks ago

Que capo

Quanta Magazine @QuantaMagazine

3 weeks ago

Que capo

6 59 270 45K 108

0 0 1 244 0

Douglas Yao @DouglasYaoDY

4 weeks ago

The most complex biological system that we meaningfully understand is a single virion. Everything more complex than that (including bacteria, single human cells, animals, and of course, humans) is totally beyond our comprehension.

19 33 201 17K 56

martin palazzo @boardsofdata

a month ago

Llego el dia de decidir entre: - Bosch - Makita - DeWalt - Black & Decker

1 0 1 316 0

François Chollet @fchollet

a month ago

Saying that deep learning is "just a bunch of matrix multiplications" is about as informative as saying that computers are "just a bunch of transistors" or that a library is "just a lot of paper and ink." It's true, but the encoding substrate is the least important part here.…

134 322 3K 206K 619

Shuangfei Zhai @zhaisf

2 months ago

Unlike an RNN, one attention block alone cannot model anything interesting. And it’s the stacking of it that does wonders. Understanding this compositionality should be at least as important as understanding the attn module itself.

36 68 953 99K 463

Download Image

Alexia Jolicoeur-Martineau @jm_alexia

2 months ago

As predicted, LLMs improves at a logarithmic rate (aka slowly). There is sadly no recursive self-improvement method that lead to AGI using these models.

0 5 18 2K 1

martin palazzo @boardsofdata

2 months ago

Mas razón aún para invertir en el modelado biomolecular de una célula: un problema abierto y complejo que de simplificarse aceleraría el desarrollo biotech (terapias, diagnóstico, etc)

Jake P. Taylor-King @wildtypehuman

2 months ago

Mas razón aún para invertir en el modelado biomolecular de una célula: un problema abierto y complejo que de simplificarse aceleraría el desarrollo biotech (terapias, diagnóstico, etc)

6 79 438 30K 247

Download Image

0 0 1 306 0

Nicholas Fabiano, MD @NTFabiano

2 months ago

Writing is thinking. Don't let AI do it all.

155 2K 13K 796K 6K

Download Image

martin palazzo @boardsofdata

2 months ago

buenas noticias para la comunidad bio-AI : empiezan a emerger los "image-nets" datasets de la biologia molecular / multi-omics que permiten hacer benchmark y medir progreso estandarizado entre distintas tecnologias openproblems.bio

0 0 5 208 0

François Fleuret @francoisfleuret

2 months ago

The lesson from the VAE is not "a VAE is just an AE with a dumb penalty" the lesson is "dumb penalties have extremely profound effects and induce incredibly sophisticated structures in deep models".

8 23 322 22K 130

Stämm @StammBio

3 months ago

1/ Can AI build a living, breathing virtual cell? Scientists are racing to create AI models that can predict how cells behave, aiming to transform biology from a lab-intensive to a computationally driven field.