Yash Dagade

I love to build :)

[Yash Dagade @yashdagad]

Sparsity is really cool (especially for JEPA based WMs) :))

[Yilun Kuang @KuangYilun]

Introducing LpWM: A Case for Sparse Representations in World Models

Dense Gaussian representations are a choice, not a requirement. We find that sparse representations can make latent dynamics easier to model for planning.

📄https://t.co/Il2krEtFi5
💻https://t.co/D1UAVtkwQD

[Yilun Kuang @KuangYilun]

Introducing LpWM: A Case for Sparse Representations in World Models

Dense Gaussian representations are a choice, not a requirement. We find that sparse representations can make latent dynamics easier to model for planning.

📄https://t.co/Il2krEtFi5
💻https://t.co/D1UAVtkwQD

[Randall Balestriero @randall_balestr]

LpWM is to LeWM what LpJEPA is to LeJEPA, another step in our search for the best latent distribution to use as target for SIGReg like objectives. Sparsity has been proving its worth again and again, but some core questions remain open...
Congrats @KuangYilun et al.!

[Rishab Jain @RishabJainK]

AI systems are already superhuman at playing games: Chess, Go, Starcraft.

@GetPrescience is the first health system that models medicine as a game. It plays to win, and it’s really good at it.

We invited a team to film a short documentary about what we’re building.

[Enigma @Enigma_AI]

Today is Drop #2: Research.

Robotics foundation models need a lot of data, and a lot of engineering even for tiny adjustments.

This feels wrong.

Today, we share with the world a sneak peek to our research - training more efficient robotic foundation models.
1/3

[Enigma @Enigma_AI]

We put 100 real AI-powered robots online.

Anyone in the world can control them right now, from a browser.

Go make one do something:

[Lucas Maes @lucasmaes_]

Would you like to join the research effort on JEPA and World Models easily?

After a full year of hard work, we’re excited to finally release stable-worldmodel:

an open-source, scalable platform built to accelerate JEPA & World Model research!

📄: https://t.co/gnxGvens5A

[Yilun Kuang @KuangYilun]

Check out this writeup about our Rectified LpJEPA paper from the NYU CDS team!

[Yash Dagade @yashdagad]

Cool post; consider giving it a read :)

https://t.co/GhDxUVq73E

[Yash Dagade @yashdagad]

Sparsity is really cool!!

[Yilun Kuang @KuangYilun]

How do we build sparsity into JEPA representations by design, while preserving task-relevant information?

Introducing Rectified LpJEPA, a JEPA architecture that learns sparse, non-negative, informative representations through principled distributional regularization. 📐

📄 Paper: https://t.co/CZtKHVvKT4
💻 Code: https://t.co/xIJMYkK9fT
📝 Blog: https://t.co/OFCl3FAmik

(1/n)

[Tim G. J. Rudner @timrudner]

Check out @KuangYilun's latest paper! Very excited about this!

We propose Rectified Distribution Matching Regularization (RDMReg) for JEPAs: a theoretically-grounded sparsification method that yields a sota sparsity–performance trade-off!

w/ @YashDagad+@randall_balestr+@ylecun

[Yilun Kuang @KuangYilun]

How do we build sparsity into JEPA representations by design, while preserving task-relevant information?

Introducing Rectified LpJEPA, a JEPA architecture that learns sparse, non-negative, informative representations through principled distributional regularization. 📐

📄 Paper: https://t.co/CZtKHVvKT4
💻 Code: https://t.co/xIJMYkK9fT
📝 Blog: https://t.co/OFCl3FAmik

(1/n)

[Tim G. J. Rudner @timrudner]

I'm so happy to share that I’ll be joining @UofT as an Assistant Professor of Statistical Sciences and Computer Science, with an appointment at the @VectorInst, in 2026!

I'm recruiting postdocs and PhD students: https://t.co/FWBh0BiDqP!

Please help me spread the word!

🧵(1/5)

[Yash Dagade @yashdagad]

layernorm is such a cool paper!

[Yash Dagade @yashdagad]

@bhuwandhingra @raghavlite This is awesome! Congrats 🎉

[Yash Dagade @yashdagad]

Alpha evolve is still shocking! It hasn’t fully sunk in yet 😳. It’s such a great time to be alive /)

[Yash Dagade @yashdagad]

@AlexBoniske This is soo cool🔥🔥👏👏

[Yash Dagade @yashdagad]

@paulg You should use that to explain entropy and how boilerplate is just low entropy stuff in life

[Yash Dagade @yashdagad]

@AlexBoniske @tejasdesigns Fire website! Stoked for the launch🔥🔥

[Yash Dagade @yashdagad]

Touch me nots are so cool