Antonin Raffin
@araffin.bsky.social
Researcher in robotics and machine learning (Reinforcement Learning). Maintainer of Stable-Baselines (SB3).
https://araffin.github.io/
https://araffin.github.io/
Pinned
Antonin Raffin
@araffin.bsky.social
· Oct 24
Post your most popular 🐦 from Twitter
Types of Reinforcement Learning Paper
Original image: @xkcd.com
Types of Reinforcement Learning Paper
Original image: @xkcd.com
Reposted by Antonin Raffin
Wow. The backlash to the 1X Neo announcement has been widespread and *merciless*.
This may be a warning to lots of humanoids companies. All your promises don’t matter to the public if your robot looks or acts dumb.
youtu.be/b_SNExtznd4?...
This may be a warning to lots of humanoids companies. All your promises don’t matter to the public if your robot looks or acts dumb.
youtu.be/b_SNExtznd4?...
Ronny Chieng Meets Neo, the World’s Stupidest Robot Maid | The Daily Show
YouTube video by The Daily Show
youtu.be
October 31, 2025 at 12:34 PM
Wow. The backlash to the 1X Neo announcement has been widespread and *merciless*.
This may be a warning to lots of humanoids companies. All your promises don’t matter to the public if your robot looks or acts dumb.
youtu.be/b_SNExtznd4?...
This may be a warning to lots of humanoids companies. All your promises don’t matter to the public if your robot looks or acts dumb.
youtu.be/b_SNExtznd4?...
Reposted by Antonin Raffin
Why self-taught engineers often outperform
michaelbastos.com/blog/why-sel...
#programming #softwaredevelopment #tech #blog
michaelbastos.com/blog/why-sel...
#programming #softwaredevelopment #tech #blog
michaelbastos.com
October 29, 2025 at 7:34 PM
Why self-taught engineers often outperform
michaelbastos.com/blog/why-sel...
#programming #softwaredevelopment #tech #blog
michaelbastos.com/blog/why-sel...
#programming #softwaredevelopment #tech #blog
Reposted by Antonin Raffin
🚨The Formalism-Implementation Gap in RL research🚨
Lots of progress in RL research over last 10 years, but too much performance-driven => overfitting to benchmarks (like the ALE).
1⃣ Let's advance science of RL
2⃣ Let's be explicit about how benchmarks map to formalism
1/X
Lots of progress in RL research over last 10 years, but too much performance-driven => overfitting to benchmarks (like the ALE).
1⃣ Let's advance science of RL
2⃣ Let's be explicit about how benchmarks map to formalism
1/X
October 28, 2025 at 1:56 PM
🚨The Formalism-Implementation Gap in RL research🚨
Lots of progress in RL research over last 10 years, but too much performance-driven => overfitting to benchmarks (like the ALE).
1⃣ Let's advance science of RL
2⃣ Let's be explicit about how benchmarks map to formalism
1/X
Lots of progress in RL research over last 10 years, but too much performance-driven => overfitting to benchmarks (like the ALE).
1⃣ Let's advance science of RL
2⃣ Let's be explicit about how benchmarks map to formalism
1/X
Reposted by Antonin Raffin
🚨 New blog post alert!
Modern package management for Robotics with Pixi!
prefix.dev/blog/reprod...
#ROS #ROSCon #ROSCon2025
Modern package management for Robotics with Pixi!
prefix.dev/blog/reprod...
#ROS #ROSCon #ROSCon2025
Pixi: Modern package management for Robotics
Developing Robots is hard; Pixi makes it easier by creating reproducible, cross-platform ROS development environments without Docker or Ubuntu lock-in.
prefix.dev
October 24, 2025 at 3:34 PM
🚨 New blog post alert!
Modern package management for Robotics with Pixi!
prefix.dev/blog/reprod...
#ROS #ROSCon #ROSCon2025
Modern package management for Robotics with Pixi!
prefix.dev/blog/reprod...
#ROS #ROSCon #ROSCon2025
Reposted by Antonin Raffin
What if we did a single run and declared victory
October 23, 2025 at 2:28 AM
What if we did a single run and declared victory
A wonderful collection of spurious correlations, correlation is not causation.
link: www.tylervigen.com/spurious-cor...
found via @stefanjudis.com newsletter
link: www.tylervigen.com/spurious-cor...
found via @stefanjudis.com newsletter
October 21, 2025 at 5:44 AM
A wonderful collection of spurious correlations, correlation is not causation.
link: www.tylervigen.com/spurious-cor...
found via @stefanjudis.com newsletter
link: www.tylervigen.com/spurious-cor...
found via @stefanjudis.com newsletter
A good video on software refactoring and redesign (about the Audacity audio editing program)
October 20, 2025 at 10:20 AM
A good video on software refactoring and redesign (about the Audacity audio editing program)
Reposted by Antonin Raffin
Video recordings of CORL 2025 talks now available! Many interesting orals / keynotes / sponsor talks / early-career talks / poster spotlights.
Day 1: www.youtube.com/watch?v=Use5...
Day 2: www.youtube.com/watch?v=rh2o...
Day 3: www.youtube.com/watch?v=9lzF...
Day 1: www.youtube.com/watch?v=Use5...
Day 2: www.youtube.com/watch?v=rh2o...
Day 3: www.youtube.com/watch?v=9lzF...
CORL 2025
YouTube video by Conference on Robot Learning
www.youtube.com
October 17, 2025 at 5:31 AM
Video recordings of CORL 2025 talks now available! Many interesting orals / keynotes / sponsor talks / early-career talks / poster spotlights.
Day 1: www.youtube.com/watch?v=Use5...
Day 2: www.youtube.com/watch?v=rh2o...
Day 3: www.youtube.com/watch?v=9lzF...
Day 1: www.youtube.com/watch?v=Use5...
Day 2: www.youtube.com/watch?v=rh2o...
Day 3: www.youtube.com/watch?v=9lzF...
Reposted by Antonin Raffin
In our little deep dive series we're now exploring how cross-compilation in the Conda ecosystem works: prefix.dev/blog/cross-c.... Back in the days, @conda-forge.org rolled this out widely to support osx-arm64 early on, and now for linux-aarch64/ppc64le.
Cross compiling in the Conda ecosystem
Cross compiling is a fundamental capability in modern software development, allowing developers to build packages for different architectures without needing access to the target hardware.
prefix.dev
October 15, 2025 at 6:11 AM
In our little deep dive series we're now exploring how cross-compilation in the Conda ecosystem works: prefix.dev/blog/cross-c.... Back in the days, @conda-forge.org rolled this out widely to support osx-arm64 early on, and now for linux-aarch64/ppc64le.
Reposted by Antonin Raffin
Rapid RL experimentation is great. But how do you catch silent errors before they slip by?
In this post, I share tools and habits that help me move quickly from idea to result without sacrificing reliability.
In this post, I share tools and habits that help me move quickly from idea to result without sacrificing reliability.
How to catch subtle RL bugs before they catch you
Tools and habits for reliable, fast RL experimentation and development
open.substack.com
October 13, 2025 at 11:29 AM
Rapid RL experimentation is great. But how do you catch silent errors before they slip by?
In this post, I share tools and habits that help me move quickly from idea to result without sacrificing reliability.
In this post, I share tools and habits that help me move quickly from idea to result without sacrificing reliability.
Reposted by Antonin Raffin
Ever since I made a video about Fourier Transforms, one of the most requested topics on the channel has been its close cousin, the Laplace Transform.
I've been having a lot of fun animating a mini-series about this topic, and the main part is now out.
youtu.be/j0wJBEZdwLs
I've been having a lot of fun animating a mini-series about this topic, and the main part is now out.
youtu.be/j0wJBEZdwLs
But what is a Laplace Transform?
YouTube video by 3Blue1Brown
youtu.be
October 12, 2025 at 12:49 PM
Ever since I made a video about Fourier Transforms, one of the most requested topics on the channel has been its close cousin, the Laplace Transform.
I've been having a lot of fun animating a mini-series about this topic, and the main part is now out.
youtu.be/j0wJBEZdwLs
I've been having a lot of fun animating a mini-series about this topic, and the main part is now out.
youtu.be/j0wJBEZdwLs
Reposted by Antonin Raffin
Updated & turned my Big LLM Architecture Comparison article into a video lecture.
The 11 LLM archs covered in this video:
1. DeepSeek V3/R1
2. OLMo 2
3. Gemma 3
4. Mistral Small 3.1
5. Llama 4
6. Qwen3
7. SmolLM3
8. Kimi 2
9. GPT-OSS
10. Grok 2.5
11. GLM-4.5/4.6
www.youtube.com/watch?v=rNlU...
The 11 LLM archs covered in this video:
1. DeepSeek V3/R1
2. OLMo 2
3. Gemma 3
4. Mistral Small 3.1
5. Llama 4
6. Qwen3
7. SmolLM3
8. Kimi 2
9. GPT-OSS
10. Grok 2.5
11. GLM-4.5/4.6
www.youtube.com/watch?v=rNlU...
The Big LLM Architecture Comparison
YouTube video by Sebastian Raschka
www.youtube.com
October 10, 2025 at 5:05 PM
Updated & turned my Big LLM Architecture Comparison article into a video lecture.
The 11 LLM archs covered in this video:
1. DeepSeek V3/R1
2. OLMo 2
3. Gemma 3
4. Mistral Small 3.1
5. Llama 4
6. Qwen3
7. SmolLM3
8. Kimi 2
9. GPT-OSS
10. Grok 2.5
11. GLM-4.5/4.6
www.youtube.com/watch?v=rNlU...
The 11 LLM archs covered in this video:
1. DeepSeek V3/R1
2. OLMo 2
3. Gemma 3
4. Mistral Small 3.1
5. Llama 4
6. Qwen3
7. SmolLM3
8. Kimi 2
9. GPT-OSS
10. Grok 2.5
11. GLM-4.5/4.6
www.youtube.com/watch?v=rNlU...
SBX (SB3 Jax) v0.23.0 is out =)!
I added CNN support for PPO.
It turns out that using a shared features extractor (CNN in this case) is important for achieving good performance on Atari games.
Perf report: wandb.ai/openrlbenchm...
github.com/araffin/sbx
I added CNN support for PPO.
It turns out that using a shared features extractor (CNN in this case) is important for achieving good performance on Atari games.
Perf report: wandb.ai/openrlbenchm...
github.com/araffin/sbx
GitHub - araffin/sbx: SBX: Stable Baselines Jax (SB3 + Jax) RL algorithms
SBX: Stable Baselines Jax (SB3 + Jax) RL algorithms - araffin/sbx
github.com
September 29, 2025 at 5:23 PM
SBX (SB3 Jax) v0.23.0 is out =)!
I added CNN support for PPO.
It turns out that using a shared features extractor (CNN in this case) is important for achieving good performance on Atari games.
Perf report: wandb.ai/openrlbenchm...
github.com/araffin/sbx
I added CNN support for PPO.
It turns out that using a shared features extractor (CNN in this case) is important for achieving good performance on Atari games.
Perf report: wandb.ai/openrlbenchm...
github.com/araffin/sbx
Training a small humanoid robot with reinforcement learning using another robot for reset.
by Kaizhe Hu et al. (ToddlerBot Stanford)
Project page: robot-trains-robot.github.io
by Kaizhe Hu et al. (ToddlerBot Stanford)
Project page: robot-trains-robot.github.io
September 29, 2025 at 8:48 AM
Training a small humanoid robot with reinforcement learning using another robot for reset.
by Kaizhe Hu et al. (ToddlerBot Stanford)
Project page: robot-trains-robot.github.io
by Kaizhe Hu et al. (ToddlerBot Stanford)
Project page: robot-trains-robot.github.io
Open-Source Hardware in the Era of Robot Learning Workshop @ CoRL 2025
Website: open-hardware-robots.github.io/CoRL2025/
Website: open-hardware-robots.github.io/CoRL2025/
September 27, 2025 at 6:19 AM
Open-Source Hardware in the Era of Robot Learning Workshop @ CoRL 2025
Website: open-hardware-robots.github.io/CoRL2025/
Website: open-hardware-robots.github.io/CoRL2025/
Reposted by Antonin Raffin
The CoRL 2025 workshop on Open-Source Hardware in the Era of Robot Learning is starting now! You can join the conversation online via live streaming: https://www.youtube.com/live/ZVPIJzF1df4
September 27, 2025 at 12:32 AM
The CoRL 2025 workshop on Open-Source Hardware in the Era of Robot Learning is starting now! You can join the conversation online via live streaming: https://www.youtube.com/live/ZVPIJzF1df4
Reposted by Antonin Raffin
📣 Call for Blog Posts at #ICLR2026 @iclr_conf
Following the success of the past iterations, we are opening the Call for Blog Posts 2026!
iclr-blogposts.github.io/2026/about/#...
Please retweet!
Following the success of the past iterations, we are opening the Call for Blog Posts 2026!
iclr-blogposts.github.io/2026/about/#...
Please retweet!
abs-0.twimg.com
September 22, 2025 at 7:44 AM
📣 Call for Blog Posts at #ICLR2026 @iclr_conf
Following the success of the past iterations, we are opening the Call for Blog Posts 2026!
iclr-blogposts.github.io/2026/about/#...
Please retweet!
Following the success of the past iterations, we are opening the Call for Blog Posts 2026!
iclr-blogposts.github.io/2026/about/#...
Please retweet!
A practical introduction to (deep) RL, providing intuitions to understand the more recent algorithms.
The plan is to start from tabular Q-learning and work our way up to Deep Q-learning (DQN). In a following post, I will continue on to Soft Actor-Critic (SAC) and its extensions.
The plan is to start from tabular Q-learning and work our way up to Deep Q-learning (DQN). In a following post, I will continue on to Soft Actor-Critic (SAC) and its extensions.
RL102: From Tabular Q-Learning to Deep Q-Learning (DQN) - A Practical Introduction to (Deep) Reinforcement Learning
araffin.github.io/post/rl102/
araffin.github.io/post/rl102/
RL102: From Tabular Q-Learning to Deep Q-Learning (DQN) | Antonin Raffin | Homepage
This blog post is meant to be a practical introduction to (deep) reinforcement learning1, presenting the main concepts and providing intuitions to understand the more recent Deep RL algorithms. For a ...
araffin.github.io
September 22, 2025 at 8:06 AM
A practical introduction to (deep) RL, providing intuitions to understand the more recent algorithms.
The plan is to start from tabular Q-learning and work our way up to Deep Q-learning (DQN). In a following post, I will continue on to Soft Actor-Critic (SAC) and its extensions.
The plan is to start from tabular Q-learning and work our way up to Deep Q-learning (DQN). In a following post, I will continue on to Soft Actor-Critic (SAC) and its extensions.
Reposted by Antonin Raffin
Next Saturday, 𝗔𝗻𝘁𝗼𝗶𝗻𝗲 𝗣𝗶𝗿𝗿𝗼𝗻𝗲 will present Pollen Robotics & Hugging Face's open-source robots, including Reachy Mini, the SO-100 arm, the Amazing Hand and the Open Duck Mini. He will discuss the sim2real challenges of making the Open Duck Mini walk, and how […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
September 21, 2025 at 12:23 PM
Next Saturday, 𝗔𝗻𝘁𝗼𝗶𝗻𝗲 𝗣𝗶𝗿𝗿𝗼𝗻𝗲 will present Pollen Robotics & Hugging Face's open-source robots, including Reachy Mini, the SO-100 arm, the Amazing Hand and the Open Duck Mini. He will discuss the sim2real challenges of making the Open Duck Mini walk, and how […]
[Original post on fosstodon.org]
[Original post on fosstodon.org]
RL102: From Tabular Q-Learning to Deep Q-Learning (DQN) - A Practical Introduction to (Deep) Reinforcement Learning
araffin.github.io/post/rl102/
araffin.github.io/post/rl102/
RL102: From Tabular Q-Learning to Deep Q-Learning (DQN) | Antonin Raffin | Homepage
This blog post is meant to be a practical introduction to (deep) reinforcement learning1, presenting the main concepts and providing intuitions to understand the more recent Deep RL algorithms. For a ...
araffin.github.io
September 18, 2025 at 3:09 PM
RL102: From Tabular Q-Learning to Deep Q-Learning (DQN) - A Practical Introduction to (Deep) Reinforcement Learning
araffin.github.io/post/rl102/
araffin.github.io/post/rl102/
Reposted by Antonin Raffin
Package building with Pixi is being rolled out! Dive into our latest blog post on crafting C++ packages.
And guess what? It’s not just for C++; Pixi plays nice with Python, Rust, ROS, Mojo, and beyond!
prefix.dev/blog/pixi-b...
And guess what? It’s not just for C++; Pixi plays nice with Python, Rust, ROS, Mojo, and beyond!
prefix.dev/blog/pixi-b...
Build C++ projects with Pixi
Painless dependency management (including shared libraries), monorepos and CI/CD is here for C++/CMake projects with Pixi.
prefix.dev
September 5, 2025 at 10:00 AM
Package building with Pixi is being rolled out! Dive into our latest blog post on crafting C++ packages.
And guess what? It’s not just for C++; Pixi plays nice with Python, Rust, ROS, Mojo, and beyond!
prefix.dev/blog/pixi-b...
And guess what? It’s not just for C++; Pixi plays nice with Python, Rust, ROS, Mojo, and beyond!
prefix.dev/blog/pixi-b...
Reposted by Antonin Raffin
bash tricks
permalink: wizardzines.com/comics/bash-...
from our zine "Bite Size Command Line": wizardzines.com/zines/bite-s...
permalink: wizardzines.com/comics/bash-...
from our zine "Bite Size Command Line": wizardzines.com/zines/bite-s...
September 3, 2025 at 7:24 PM
bash tricks
permalink: wizardzines.com/comics/bash-...
from our zine "Bite Size Command Line": wizardzines.com/zines/bite-s...
permalink: wizardzines.com/comics/bash-...
from our zine "Bite Size Command Line": wizardzines.com/zines/bite-s...
Reposted by Antonin Raffin
This is absolutely true -- this is a superb and much-needed consolidation of so much of modern RL. Kevin, inquiring minds want to understand the process you use to put this artwork together! @sirbayes.bsky.social Perhaps this is also the ultimate benchmark for Gemini Deep Research reports. ;-p
Reminded again of Kevin Murphy's excellent RL overview: arxiv.org/abs/2412.05265
A lot of the stuff covered here really is at the cutting edge and not compiled so nicely anywhere else
A lot of the stuff covered here really is at the cutting edge and not compiled so nicely anywhere else
Reinforcement Learning: An Overview
This manuscript gives a big-picture, up-to-date overview of the field of (deep) reinforcement learning and sequential decision making, covering value-based methods, policy-based methods, model-based m...
arxiv.org
September 3, 2025 at 4:41 AM
This is absolutely true -- this is a superb and much-needed consolidation of so much of modern RL. Kevin, inquiring minds want to understand the process you use to put this artwork together! @sirbayes.bsky.social Perhaps this is also the ultimate benchmark for Gemini Deep Research reports. ;-p
Reposted by Antonin Raffin
Weekend project: building a (site) search engine www.redblobgames.com/blog/2025-08... just for fun! :)
Let’s write a search engine, part 1 of 2
www.redblobgames.com
September 1, 2025 at 2:48 AM
Weekend project: building a (site) search engine www.redblobgames.com/blog/2025-08... just for fun! :)