What do those representations look like? Each color corresponds to a different task. Predictive representations lead to structured, high-dim latent spaces. W/out them, representations collapse & are less expressive. 9/🧵
Why does it work? Our analyses suggest predictive objectives lead to: 💰Richer latent representations 💠Higher effective representation dimensionality 🧠Fewer dormant neurons ⚖️More stable value learning 8/🧵
Users share the arXiv paper and GitHub code on predictive objectives in RL while thanking readers and crediting the lead author plus co-authors for the collaborative work.
No Digg Deeper questions have been answered for this story yet.
Finally, most of the credit goes to @johanobandoc who led this project! Thanks to a great set of co-authors @luli_airl, Scott Fujimoto, @pierrelux, and @AaronCourville for countless insightful discussions throughout this work.
Paper: https://arxiv.org/abs/2606.05555 Code: https://github.com/johanobandoc/ScaleMRL Thanks for reading! If you're interested in multitask RL, representation learning, or scalable reinforcement learning, we'd love to hear your thoughts and feedback. I'll also be at #ICML2026 next week 🇰🇷
Perhaps even more interesting: The learned representations transfer remarkably well. MR.Q achieves stronger zero-shot performance on unseen tasks and adapts faster during few-shot finetuning, suggesting that better representations naturally lead to better transfer. 10/🧵
What do those representations look like? Each color corresponds to a different task. Predictive representations lead to structured, high-dim latent spaces. W/out them, representations collapse & are less expressive. 9/🧵
The broader takeaway is not that planning is unnecessary. Rather, we should carefully distinguish the benefits of representation learning from the benefits of planning. As RL continues to scale, understanding which ingredients matter is essential Paper and code are below. 👇
Perhaps even more interesting: The learned representations transfer remarkably well. MR.Q achieves stronger zero-shot performance on unseen tasks and adapts faster during few-shot finetuning, suggesting that better representations naturally lead to better transfer. 10/🧵