Pablo Samuel Castro's Avatar

Pablo Samuel Castro

@pcastr.bsky.social

SeΓ±or swesearcher @ Google DeepMind, adjunct prof at UniversitΓ© de MontrΓ©al and Mila. Musician. From πŸ‡ͺπŸ‡¨ living in πŸ‡¨πŸ‡¦. https://psc-g.github.io/

3,566 Followers  |  343 Following  |  360 Posts  |  Joined: 19.11.2024
Posts Following

Posts by Pablo Samuel Castro (@pcastr.bsky.social)

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

26.02.2026 21:17 β€” πŸ‘ 10    πŸ” 3    πŸ’¬ 0    πŸ“Œ 1

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!πŸ‘‡πŸ½

17.02.2026 11:08 β€” πŸ‘ 18    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
The Formalism-Implementation Gap in Reinforcement Learning Research The last decade has seen an upswing in interest and adoption of reinforcement learning (RL) techniques, in large part due to its demonstrated capabilities at performing certain tasks at "super-human l...

This is something I talk about in my paper, where I suggest being explicit about {\gamma}_train (some methods use multiple gammas during training) and \gamma_eval.
One of my students is empirically investigating this and, as one would expect, it can have a huge impact.

arxiv.org/abs/2510.16175

29.01.2026 10:08 β€” πŸ‘ 12    πŸ” 1    πŸ’¬ 1    πŸ“Œ 1
Post image

Today's wordle spoke volumes.

Wordle 1,677 2/6

β¬›πŸŸ©β¬›πŸŸ©πŸŸ©
🟩🟩🟩🟩🟩

P.S. here's a selfie of my x-country ski commute this morning πŸ˜ƒπŸŽΏ

21.01.2026 14:55 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Today's wordle energized me

Wordle 1,662 2/6

πŸŸ¨πŸŸ©πŸŸ©β¬›β¬›
🟩🟩🟩🟩🟩

07.01.2026 04:19 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image

How I started 2026, hoping to keep this spirit up throughout the year!
Happy new year!

02.01.2026 20:37 β€” πŸ‘ 10    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Another one of my favorite spots in Ecuador: El PailΓ³n del Diablo πŸ‡ͺπŸ‡¨

30.12.2025 23:23 β€” πŸ‘ 5    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Gets better the closer you get πŸ”οΈπŸ‡ͺπŸ‡¨

29.12.2025 22:23 β€” πŸ‘ 12    πŸ” 0    πŸ’¬ 0    πŸ“Œ 1

Quito, Ecuador

26.12.2025 17:53 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image

Will never tire of this view ❀️πŸ‡ͺπŸ‡¨

26.12.2025 14:26 β€” πŸ‘ 13    πŸ” 0    πŸ’¬ 1    πŸ“Œ 1
RLJ | RLC Call for Papers

Hi RL Enthusiasts!

RLC is coming to Montreal, Quebec, in the summer: Aug 16–19, 2026!

Call for Papers is up now:
Abstract: Mar 1 (AOE)
Submission: Mar 5 (AOE)

Excited to see what you’ve been up to - Submit your best work!
rl-conference.cc/callforpaper...

Please share widely!

23.12.2025 22:16 β€” πŸ‘ 61    πŸ” 28    πŸ’¬ 0    πŸ“Œ 7

Today's wordle made me feel warm.

Wordle 1,646 2/6

β¬›πŸŸ¨πŸŸ©β¬›β¬›
🟩🟩🟩🟩🟩

22.12.2025 04:19 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Despite some of its flaws, OpenReview has had a tremendously positive impact on our research community.
This is a great initiative! Join me, and many others, in donating to ensure OpenReview not only remains a positive force, but is able to improve.

20.12.2025 21:40 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

It was really great to chat with so many friends that I don't always get a chance to see, and to meet new ones, especially during our morning runs.
I really enjoyed this #NeurIPS2025 , but am also exhausted πŸ₯°πŸ«©

Until next time!

bsky.app/profile/did:...

10.12.2025 21:48 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Finally, on Saturday I gave a talk at the Embodied World Models for Decision Making workshop on our recent paper on using foundation models to automatically design Reward Machines for improved RL training:

Paper: arxiv.org/abs/2510.14176

Post:
bsky.app/profile/roge...

10.12.2025 21:48 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image

On Friday afternoon Eduardo Pignatelli and I *should have* presented our paper on NAVIX: Jax-based MiniGrid for super speedy experiments! πŸš€
Due to our miscommunication our poster was MIA πŸ˜…

openreview.net/forum?id=lPV...

10.12.2025 21:48 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image Post image

Also on Friday morning, Jiashun and @johanobandoc.bsky.social presented our Measure Gradients, not Activations paper: moving from ReDo to ReGraMa to improve learning performance of RL agents!

openreview.net/forum?id=FjN...

10.12.2025 21:48 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image Post image Post image Post image

On Friday morning @roger-creus.bsky.social @johanobandoc.bsky.social @glenberseth.bsky.social and I presented our (spotlight) Stable Gradients paper: scaling depth and width of RL networks by avoiding vanishing gradients!
(Also presented this at LXAI workshop)

openreview.net/forum?id=Vqj...

10.12.2025 21:48 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image Post image

On Thursday Ghada Sokar and I presented our Mind The Gap paper: better understanding of (and simplifying) how to scale RL networks!

openreview.net/forum?id=LrB...

10.12.2025 21:48 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image Post image

On Wednesday Reggie and Evangelos presented our MetaWorld+ paper: Multi-task continuous control done right!

openreview.net/forum?id=1de...

10.12.2025 21:48 β€” πŸ‘ 2    πŸ” 1    πŸ’¬ 1    πŸ“Œ 0

This #NeurIPS2025 was tiring, but it was fantastic to connect with so many friends and colleagues!
I was so busy I didn't get a chance to tweemote our papers at the conference, so I'll remedy that with this post-hoc thread: πŸ‘‡πŸΎ

10.12.2025 21:48 β€” πŸ‘ 2    πŸ” 1    πŸ’¬ 1    πŸ“Œ 0

Sixth, and last, #runconference at #NeurIPS2025 had the best turnout yet!
Thanks to everyone who came out, until the next conference!
The second picture is of those who stayed until the end πŸ˜…πŸ…
πŸ€–πŸƒπŸΎπŸ‘‹πŸΎ

07.12.2025 17:49 β€” πŸ‘ 11    πŸ” 0    πŸ’¬ 0    πŸ“Œ 1

Fifth #runconference at #NeurIPS2025 , good turnout again!
Tomorrow is the last run (at least with me), so if you've had FOMO this week, you have one more chance!
πŸ€–πŸƒπŸΎ

07.12.2025 00:11 β€” πŸ‘ 7    πŸ” 0    πŸ’¬ 0    πŸ“Œ 1

Fourth #runconference at #NeurIPS2025 had the best turnout yet!
Help us beat it tomorrow, same place, same time (follow tweet thread for details)!
πŸ€–πŸƒπŸΎ

06.12.2025 01:22 β€” πŸ‘ 13    πŸ” 0    πŸ’¬ 0    πŸ“Œ 1

Third #runconference at #NeurIPS2025 was great, including special guest @JeffDean !
Same place, same time, tomorrow morning if you want to join!
πŸ€–πŸƒπŸΎ
(I forgot to post yesterday's pic, so including it here)

04.12.2025 19:15 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 1

. @roger-creus.bsky.social and @johanobandoc.bsky.social did an awesome job presenting our paper at the LatinX in AI workshop at #NeurIPS .

If you missed it, come to our poster #310 on Friday 11am, exhibit hall C,D,E.

03.12.2025 00:53 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Join us tomorrow at #runconference for #NeurIPS2015, same time, same place (see quoted thread for details)!

It was three of us this morning but I forgot to take a picture before we separated, so here's a selfie to prove we actually went out 🀭

Hope to see more of you tomorrow!

02.12.2025 21:38 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 1
Post image

Just landed in San Diego for #NeurIPS2025 !
If anyone wants to join me for #runconference tomorrow, meet me at the pin at 7:00am. We'll play distance and pace by ear, based on who shows up
See you then!
πŸ€–πŸƒπŸΎ
maps.app.goo.gl/W7fedHdpsrFn...

01.12.2025 22:17 β€” πŸ‘ 5    πŸ” 0    πŸ’¬ 1    πŸ“Œ 1

Today's wordle was received as a gift.

Wordle 1,622 2/6

β¬›πŸŸ©πŸŸ©β¬›β¬›
🟩🟩🟩🟩🟩

28.11.2025 04:15 β€” πŸ‘ 6    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Post image

My lab is looking for new students who are very passionate about foundational models and planning/RL/robotics. Apply via Mila. I will also be at #NeurIPS to discuss research ideas and opportunities. See notes below for application advice.

19.11.2025 15:10 β€” πŸ‘ 17    πŸ” 4    πŸ’¬ 2    πŸ“Œ 1