Please check out our #ICML2026 🇰🇷 poster for ReMax tomorrow, where we show how exploration can emerge in RL by retrying under uncertainty.
Long road since this project started in 2022 😂 Huge thanks to all co-authors, especially @PaavoParmas and first author @nissymori1 🙏
Our paper "Emergence of Exploration in Policy Gradient Reinforcement Learning via Retrying" is accepted at #ICML2026 🇰🇷
Proposed a novel exploration objective called ReMax, evaluating best of multiple trials under uncertainty.
The objective comes from the basic question,
Why do RL agents need to explore?
We argue it is because
♻️ Agents are allowed to retry (otherwise, the rational choice is the current best action).
📈 Return is uncertain (otherwise, no point in trying suboptimal actions.)
ReMax naturally captures these intuitions by modeling the distribution of returns and evaluating the maximum over multiple retries, thereby encouraging agents to select actions that are currently suboptimal but highly uncertain.
The diagram is inspired by the Vector Policy Optimization (VPO) paper.
🧵1/n
Our paper "Emergence of Exploration in Policy Gradient Reinforcement Learning via Retrying" is accepted at #ICML2026 🇰🇷
Proposed a novel exploration objective called ReMax, evaluating best of multiple trials under uncertainty.
The objective comes from the basic question,
Why do RL agents need to explore?
We argue it is because
♻️ Agents are allowed to retry (otherwise, the rational choice is the current best action).
📈 Return is uncertain (otherwise, no point in trying suboptimal actions.)
ReMax naturally captures these intuitions by modeling the distribution of returns and evaluating the maximum over multiple retries, thereby encouraging agents to select actions that are currently suboptimal but highly uncertain.
The diagram is inspired by the Vector Policy Optimization (VPO) paper.
🧵1/n
⛳ Introducing purejaxgcrl: goal-conditioned reinforcement learning in end-to-end JAX for discrete action environments!
Train generalist, goal-conditioned agents in minutes on a single GPU.
1/
Ludax is inspired by both the Ludii game description language (https://t.co/DQW12FjUZp) and the excellent PGX library of board game implementations in JAX (https://t.co/v0COK0TEsY)
Our goal is to combine the best parts of both projects: flexibility and speed!