Pablo Samuel Castro
@pcastr.bsky.social
Señor swesearcher @ Google DeepMind, adjunct prof at Université de Montréal and Mila. Musician. From 🇪🇨 living in 🇨🇦. https://psc-g.github.io/
Fifth and last #runconference at #ICML2026 was great! Thanks to everyone who came out, and until next time!
Pablo Samuel Castro@pcastr.bsky.social · 4w agoFourth #runconference at #ICML2026 was a bit of a bust, as separate running club cancelled last minute 🫤 Still had a nice (and very humid) solo run. Tomorrow will be the last one, we'll meet at 8am at Hannam station for a run around the river, join us! naver.me/F5swAa7n
Fourth #runconference at #ICML2026 was a bit of a bust, as separate running club cancelled last minute 🫤 Still had a nice (and very humid) solo run. Tomorrow will be the last one, we'll meet at 8am at Hannam station for a run around the river, join us! naver.me/F5swAa7n
Pablo Samuel Castro@pcastr.bsky.social · 4w agoNo rain for third #runconference at #ICML2026 but it was quite humid! Tomorrow we'll join a separate running club at 8am, running around some of the main sights in Seoul, join us! Apparently anyone can join, just register here: luma.com/o0wqzdyc
Interested in the differences in strategic behavior between humans and LLMs? Come chat with me and @caroline-wang.bsky.social this Thursday afternoon, 5pm, #ICML2026 poster session in Hall A (Poster #401)!
Caroline Wang@caroline-wang.bsky.social · 6mo ago[1/n] Just wrapped up 7 months interning with @pcastr.bsky.social at Google DeepMind and I'm so excited to share our work: arxiv.org/abs/2602.10324. TLDR: We used LLM-powered program synthesis to automatically model and discover differences between human and LLM strategic behavior
No rain for third #runconference at #ICML2026 but it was quite humid! Tomorrow we'll join a separate running club at 8am, running around some of the main sights in Seoul, join us! Apparently anyone can join, just register here: luma.com/o0wqzdyc
Pablo Samuel Castro@pcastr.bsky.social · 4w agoSecond #runconference at #ICML2026 was *very* wet, but we did it! Join us tomorrow, same place and same time!
Second #runconference at #ICML2026 was *very* wet, but we did it! Join us tomorrow, same place and same time!
Pablo Samuel Castro@pcastr.bsky.social · 4w agoFirst #runconference at #ICML2026 was rainy but good! We ran up to the city wall and got nice views of the city. Join us tomorrow, same place, same time!
Today I'll be presenting this spotlight poster with Ali at 2:30-4:15, in Hall A, poster #303. Come chat with us about good representations for RL! #ICML2026
Pablo Samuel Castro@pcastr.bsky.social · 5mo agoNew paper 🚨 "Stable Deep Reinforcement Learning via Isotropic Gaussian Representations" Deep RL suffers from unstable training, representation collapse, and neuron dormancy. We show that a simple geometric insight, isotropic Gaussian representations, can fix this. Here's how 👇
Today at the #ICML2026 LatinX in AI workshop Ali did a great job presenting our paper! We'll be presenting the poster at the main conference, tomorrow Wednesday July 8th, 2:30-4:15, Hall A, poster #303. Come chat with Ali and me then!
Pablo Samuel Castro@pcastr.bsky.social · 5mo agoNew paper 🚨 "Stable Deep Reinforcement Learning via Isotropic Gaussian Representations" Deep RL suffers from unstable training, representation collapse, and neuron dormancy. We show that a simple geometric insight, isotropic Gaussian representations, can fix this. Here's how 👇
First #runconference at #ICML2026 was rainy but good! We ran up to the city wall and got nice views of the city. Join us tomorrow, same place, same time!
Pablo Samuel Castro@pcastr.bsky.social · last mo.At Seoul for #ICML2026 and want to start your days with a run? Meet me tomorrow Tuesday at 7am in front of COCODA [NAVER Maps] COCODA 398-6 Jongno 5(o)-ga Jongno-gu Seoul naver.me/xuFb2ppi #runconference
At Seoul for #ICML2026 and want to start your days with a run? Meet me tomorrow Tuesday at 7am in front of COCODA [NAVER Maps] COCODA 398-6 Jongno 5(o)-ga Jongno-gu Seoul naver.me/xuFb2ppi #runconference
I'm excited to share our latest work “Representation Learning Enables Scalable Multitask Deep Reinforcement Learning”. In this work, we revisit a fundamental question in reinforcement learning: What if representation learning was the key ingredient behind scalable RL? 1/🧵
really proud of this work done with fantastic collaborators, in particular daniel, @kevinjmiller.bsky.social & @neurokim.bsky.social ! i learned so much about neuroscience, & it was great to work on using AI techniques to advance science. read kevin's thread below or, even better, read the paper!
Kevin J Miller@kevinjmiller.bsky.social · 2mo agoComputational models are a key part of science but discovering new ones is hard! DataDIVER discovers concise models from data, which surface new mechanistic ideas and clear predictions for future experiments From Google Deepmind Neuroscience Lab + collaborators www.biorxiv.org/content/10.6...
I wish there were a way to increase diversity in workshop keynotes/panelists. There are a few famous researchers who end up being keynotes/panelists on multiple workshops, which means lots of other great researchers are not getting those opportunities.
Fun last #runconference at #ICLR2026, included sidewalk and beach running, a little soccer ball kicking, and going in the Ocean! also passed some lifeguards in training during my last sprint
Pablo Samuel Castro@pcastr.bsky.social · 3mo agoAnother great #runconference this morning at #ICLR2026 ! Tomorrow is my last day so trying something different: BEACH RUN Let's meet at the pin below at 7:30am, run for a bit, then go in the water for a swim! maps.app.goo.gl/KZfuFtgsdmn4...
Another great #runconference this morning at #ICLR2026 ! Tomorrow is my last day so trying something different: BEACH RUN Let's meet at the pin below at 7:30am, run for a bit, then go in the water for a swim! maps.app.goo.gl/KZfuFtgsdmn4...
Pablo Samuel Castro@pcastr.bsky.social · 3mo agoGreat first #runconference at #ICLR2026 ! Join us tomorrow at the same place, although we'll start at 6:30am to avoid the heat 🥵
I will be giving a talk about this work at the "From Human Cognition to AI Reasoning: Models, Methods, and Applications" #ICLR2026 workshop 🗓️Sunday, April 26, 2026 10:30 AM BRT 📍 Room 202 C As well as a poster at 3:30pm in the same room. Come by!
Caroline Wang@caroline-wang.bsky.social · 6mo ago[1/n] Just wrapped up 7 months interning with @pcastr.bsky.social at Google DeepMind and I'm so excited to share our work: arxiv.org/abs/2602.10324. TLDR: We used LLM-powered program synthesis to automatically model and discover differences between human and LLM strategic behavior
Still at #ICLR2026 and still into RL? Come talk to @johanobandoc.bsky.social @waltermayor.bsky.social and me at our poster today (Saturday)! 🗓️April 25, 3:15 PM -5:45 PM BRT 📍Pavilion 4· #️⃣ Poster #4515
Pablo Samuel Castro@pcastr.bsky.social · 10mo ago🔊Simplicial Embeddings (SEMs) Improve Sample Efficiency in Actor-Critic Agents🔊 In our recent preprint we demonstrate that the use of well-structured representations (SEMs) can dramatically improve sample efficiency in RL agents. 1/X
Great first #runconference at #ICLR2026 ! Join us tomorrow at the same place, although we'll start at 6:30am to avoid the heat 🥵
Pablo Samuel Castro@pcastr.bsky.social · 3mo agoAre you at #ICLR2026 and staying at (or near) the Marriott? Come run with me tomorrow (Friday) at 7am! Starting from the Marriott. #runconference maps.app.goo.gl/zBzQjRLDMWHi...
Are you at #ICLR2026 and staying at (or near) the Marriott? Come run with me tomorrow (Friday) at 7am! Starting from the Marriott. #runconference maps.app.goo.gl/zBzQjRLDMWHi...
We have the keynote speakers for RLC2026 now! Thrilled to welcome Rika Antonova, Sheila McIlraith, Marc G. Bellemare, Danijar Hafner, Balaraman Ravindran! Details: rl-conference.cc/index.html The RL community is coming together this August in Montréal, Québec, Canada. Hope you make it!
RLC 2026
rl-conference.cc
Applications for the global Google PhD Fellowship Program are open! Fellowships directly support grad students doing exceptional and innovative research in computer science and related fields as they pursue their PhD. Learn more and apply by April 30 at goo.gle/phdfellowship.
Google PhD fellowship program
The Google PhD Fellowship Program recognizes outstanding graduate students doing exceptional work in computer science, related disciplines, or promising research areas.
goo.gle
Exploring SigReg regularization for RL has been on my "todo when I have time" list for a while. While I haven't fully read this paper yet, it appears to do a fantastic thorough investigation, and it looks quite promising!
Pablo Samuel Castro@pcastr.bsky.social · 5mo agoNew paper 🚨 "Stable Deep Reinforcement Learning via Isotropic Gaussian Representations" Deep RL suffers from unstable training, representation collapse, and neuron dormancy. We show that a simple geometric insight, isotropic Gaussian representations, can fix this. Here's how 👇
New paper 🚨 "Stable Deep Reinforcement Learning via Isotropic Gaussian Representations" Deep RL suffers from unstable training, representation collapse, and neuron dormancy. We show that a simple geometric insight, isotropic Gaussian representations, can fix this. Here's how 👇
In light of the ongoing conflict in the Middle East, RLC decided to remove the abstract deadline: rl-conference.cc/callforpaper... The only deadline is for the full paper: Mar 5(AOE) openreview.net/group?id=rl-... Affected folks may also contact the PCs to discuss deadline extensions before Mar 5.
RLC 2026 Conference
Welcome to the OpenReview homepage for RLC 2026 Conference
openreview.net
RLC 2026 Call for Workshop is live on OpenReview! Submission deadline: Mar 12 (AoE). Full details here: rl-conference.cc/call_for_wor... @glenberseth.bsky.social @eugenevinitsky.bsky.social @twkillian.bsky.social @schaul.bsky.social @sologen.bsky.social @audurand.bsky.social @bradknox.bsky.social
RLJ | RLC Call for Workshops
rl-conference.cc
Reinforcement Learning Conference@rl-conference.bsky.social · 5mo agoRLC2026 Call for Workshops! We’re already live: openreview.net/group?id=rl-... Here's the opportunity to help shape the conference & spotlight your own RL focus areas. Call: rl-conference.cc/call_for_wor... Deadline: Mar 12 (AoE) And don't forget the awesome banquet :) www.cirquedusoleil.com/ludo
It's been such a pleasure working with @caroline-wang.bsky.social the last few months. She is a fantastic researcher, engineer, & person, and I really hope we can work together again. Check out the paper that came out of her time with us, comparing strategic behavior of humans and LLMs!👇🏽
Caroline Wang@caroline-wang.bsky.social · 6mo ago[1/n] Just wrapped up 7 months interning with @pcastr.bsky.social at Google DeepMind and I'm so excited to share our work: arxiv.org/abs/2602.10324. TLDR: We used LLM-powered program synthesis to automatically model and discover differences between human and LLM strategic behavior
Today's wordle spoke volumes. Wordle 1,677 2/6 ⬛🟩⬛🟩🟩 🟩🟩🟩🟩🟩 P.S. here's a selfie of my x-country ski commute this morning 😃🎿
Today's wordle energized me Wordle 1,662 2/6 🟨🟩🟩⬛⬛ 🟩🟩🟩🟩🟩
