Mateus Machado

Master and Ph.D. candidate in computer science on Universidade Federal de Pernambuco. Focus on Reinforcement Learning.

[Shuai Bai @shuai_bai_]

Excited to share Qwen-VLA paper, our exploration of generalist Vision-Language-Action models.
It extends Qwen’s multimodal backbone from visual understanding and reasoning to continuous action generation and trajectory prediction.
Paper:
https://t.co/9jvRW0nI8B

[Dante Raimundo 🇻🇪 @RaimundoDante]

Quem faz propaganda de bet deveria ser proibido por lei falar mal do Bolsa Família.

De nada.

[Mateus Machado @MateusGM]

Quando eu tiver minha fazendinha afastada, vou criar UM porquinho e vou chamar ele de baconzitos

[老季 @XiaoJi0403]

猪其实是特别特别特别特别爱干净的动物，他们只是被关在猪圈没有条件天天洗澡

[Matheus @matheustuite]

A idade adulta é aguentar vinte problemas sérios em silêncio e depois colapsar porque caiu uma colher.

[Luiz-Eduardo Del-Bem @dudelbem]

Ironicamente, este penduricalho sozinho é maior que o salário inicial de um professor de universidade federal.

[Antonio Li @AntonioSitongLi]

Introducing NORI L1.

The best robot under $1,000.

Pre-order today. Ships summer 2026.

[Mateus Machado @MateusGM]

@tom_doerr Man, i just keep saving your posts.
Great job!

[Iranian Force @MrImranPk]

Retweet if you believe Netanyahu is a war criminal.

[Richard Sutton @RichardSSutton]

In war, both sides lose. That we don’t learn this is the greatest tragedy.

[Yagu @yagogmrs]

eu indo comprar chocolate com dinheiro que eu achei na rua enquanto o meu avô e o resto da minha família morre de fome em casa comendo repolho e epa

[Mateus Machado @MateusGM]

I see sentdex content since 2018, and this guy still does very cool stuff!

[Harrison Kinsley @Sentdex]

Introducing KimoLab: Kimodo + mjlab for prompt to physics-based motion matching for the unitree g1.

I accidentally had claude make this and it's actually pretty cool.

[gabigolo @gabigolu]

Que isso cara kkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk

[Mateus Machado @MateusGM]

@antigravity Plus, I made a PnP RL-env using it with robotiq gripper!

[Mateus Machado @MateusGM]

Hey guys!
Yesterday I was talking to my last job friends and remembered that we had a DENSO VS050 model.
They also increased some cool stuff on controlling via ROS2.
So I thought "well.. why not doing it on mujoco?"
Also, I wanted to test @antigravity 😂
https://t.co/KSRgRx4G6g

[ducy @destroynseduce]

e pensar q eu quase perdi dinheiro

se não aguenta nem lava imagina uma coquinha 0 gelada

[Clayson Felizola @claysonfelizola]

Este aqui é o melhor vídeo do Bento Ribeiro de todos os tempos ou ainda tem outro?

[Harrison Kinsley @Sentdex]

In a world of PPO everything for reinforcement learning, I've been tinkering with SAC for training a quadruped gait.

This gait is trained purely on CPU (training on one of the Dell GB10s) on a single environment. Training any particular run is obviously slower than PPO on an RTX Pro 6000 with 8092 envs, if you already know the exact hyperparams/rwd function for your PPO algo... but, if we're honest with ourselves, then we know we usually spend days tuning our PPO algo and fighting it to do what we want.

In contrast, SAC has kind of been a breath of fresh air, very amenable to changing the reward function to tune behavior. So far, my first attempts to tune things have consistently just worked immediately rather than 15 different variations of reward hacking only to find previous tuned behaviors got lost in the process. There is also FastSAC, which I've not yet tried, but can speed things up potentially and introduce scale back into the equation.

My main painpoint in getting SAC to work for gait was actually getting it to learn to step. It seems as though SAC is not as good as PPO at significant exploration on its own. I ended up starting with a sinusoidal gait (basically just a rule to make legs swing) as training wheels then blended it out through training as phase 1, then began working on smoothing things out after this.

I think if we look at end to end dev time rather than any particular run that finally managed to work, SAC may actually be the "faster" algorithm to train. Quadruped gaits are inherently easier than bipedal and maybe there are areas where SAC falls short, but I'll definitely be spending more time with SAC.

[Harrison Kinsley @Sentdex]

the high of getting hardware to move in the real world based on algos you write is the greatest drug known to man

[Richard Sutton @RichardSSutton]

I stand with Venezuela.

[Massimo @Rainmaker1973]

She's lucky to live in the 21st century, not in the 15th century after this