The paper investigates when intrinsic rewards effectively drive exploration in reinforcement learning. It introduces a formal criterion that evaluates policies based on the counterfactual information they acquire, comparing how well their histories can replace experience from alternative policies. Using a simple environment, the authors show that common intrinsic reward objectives—count-based, prediction-error, empowerment, and information-gain—can lead to Pareto-suboptimal exploration under this criterion, and they propose conditions and a new objective that better align with optimal exploration.
By Scott W. Viteri (Stanford University), Laura Gomezjurado Gonzalez (Stanford University), Clark Barrett (Stanford University)
arXiv:2605.14211v4 Announce Type: replace
Abstract: Long-horizon visuomotor tasks remain a fundamental challenge in AI, as current methods rely on hand-engineered rewards or action-labeled demonstrat...
By Benjamin Schneider, Xavier Schneider, Victor Zhong, Sun Sun
arXiv:2606. 19476v1 Announce Type: cross Abstract: Effective machine learning depends not only on how we model data, but also on what data we choose to collect.
By Eric Elmoznino, Sangnie Bhardwaj, Johannes von Oswald, Rajai Nasser, Blaise Ag\"uera y Arcas, Jo\~ao Sacramento, Rif A. Saurous, Guillaume Lajoie
arXiv:2609.38334v1 Announce Type: new
Abstract: Large language models (LLMs) are increasingly deployed as agents for multi-step decision-making, yet transfer poorly to unseen environments. World-mode...
By Yuhan Guo, Jinming Liu, Liang Xu, Ziqiang Li, Jianguo Huang, Zhicheng Wang, Hu Zhu, Qiuyu Chen, Yuntao Wei, Xin Jin, Wenjun Zeng
arXiv:2507.11482v5 Announce Type: replace
Abstract: Artificial learning systems are graduating from passive learners to increasingly autonomous agents, lending pragmatic urgency to the question of wh...
By Mani Hamidi, Terrence W. Deacon
arXiv:2605. 14211v3 Announce Type: replace Abstract: Long-horizon embodied tasks remain a fundamental challenge in AI, as current methods rely on hand-engineered rewards or action-labeled demonstrations, neither of which scales.
By Benjamin Schneider, Xavier Schneider, Victor Zhong, Sun Sun
Effective machine learning depends not only on how we model data, but also on what data we choose to collect. While large sequence models have revolutionized data modeling, the problem of automated data selection, or "intrinsic curiosity", remains a significant challenge.
arXiv:2607. 29419v1 Announce Type: cross Abstract: In reinforcement learning, exploration with sparse and delayed rewards presents a significant challenge due to the limited feedback available for guiding the learning process.
By Bumgeun Park, Donghwan Lee
arXiv:2609.36473v1 Announce Type: new
Abstract: Temporal abstraction via options can improve exploration in large environments. However, existing option discovery algorithms find subgoals that target...
By Akhil Bagaria, Anita De Mello Koch, George Konidaris
arXiv:2606. 17680v1 Announce Type: new Abstract: Reinforcement learning (RL) has emerged as a powerful paradigm for training Large Language Models (LLMs) as agents.
By Zhitong Wang, Songze Li, Hao Peng, Shuzheng Si, Yi Wang, Maosong Sun, Juanzi Li
PhantomEnvironments is a framework that trains large language model agents in synthetic, rule‑generated fictional worlds. By creating multi‑turn reinforcement learning environments where agents search templated articles to answer multi‑hop questions, the approach eliminates the need for costly human data or hallucinated LLM‑generated settings. Agents trained in these zero‑cost, purely rule‑based worlds transfer effectively to real‑world multi‑hop search benchmarks, often surpassing models trained on real data, and demonstrate scalable search behavior that grows linearly with question difficulty.
By Anmol Kabra, Swathi Saravana Selvam, Albert Gong, Chao Wan, Christian Belardi, Dongyoung Go, Katie Z. Luo, Kilian Q. Weinberger
The paper introduces two algorithms for personalized federated reinforcement learning: PF-UCBVI, which achieves optimal linear speedup by separating shared dynamics from individual goals, and PF-EFVI, a greedy method that uses reward diversity to guarantee state-action coverage without explicit exploration bonuses. The authors prove that PF-EFVI attains logarithmic regret under sufficient reward diversity, showing that disagreement among agents can shift the exploration burden from temporal to spatial aspects. This demonstrates that heterogeneous rewards can be leveraged as a structural asset to facilitate efficient collective learning.
By Jiin Woo, Gauri Joshi, Yuejie Chi