The paper introduces Reward Ensemble under Confidence (REC), a probabilistic reward learning framework for preference-based reinforcement learning that models per‑timestep reward uncertainty using an ensemble of distributional reward models. REC incorporates uncertainty into the preference loss and uses model disagreement to drive exploration, achieving 88.4% of shaped‑reward performance on acrobatic quadrotor control versus 55.2% with standard Preference PPO. The authors train policies in simulation and transfer them zero‑shot to real quadrotors, demonstrating complex acrobatic maneuvers learned solely from human preference feedback, and validate REC on a continuous‑control benchmark.
By Colin Merk, Ismail Geles, Jiaxu Xing, Angel Romero, Giorgia Ramponi, Davide Scaramuzza
arXiv:2506. 13741v2 Announce Type: replace-cross Abstract: Preference-based reinforcement learning (PbRL) has emerged as a promising approach for learning behaviors from human feedback without predefined reward functions.
By Brahim Driss, Alex Davey, Riad Akrour
arXiv:2604. 26360v2 Announce Type: replace-cross Abstract: Reinforcement learning from human feedback (RLHF) systems face a compounding alignment challenge: not only are learned reward models uncertain about unseen state-action pairs, but the human preference annotations they are trained on are themselves inconsistent, context-dependent, and noisy.
By Disha Singha
arXiv:2606. 00367v1 Announce Type: cross Abstract: Reinforcement learning problems typically define the goal as maximizing the expected value of a scalar reward function.
By Jonathan Cola\c{c}o Carr, Prakash Panangaden, Doina Precup, Benjamin Van Roy
arXiv:2609.38938v1 Announce Type: new
Abstract: Reinforcement learning with human feedback (RLHF) learns from human comparisons, which can be corrupted or deliberately manipulated. This paper studies...
By Xinyi Ni, Lifeng Lai
The paper introduces BUMEX, a reinforcement learning exploration strategy that leverages a set of prior models containing the true transition kernel and reward function. By optimizing over this model set, the method derives upper and lower bounds on the Q‑function to guide exploration, providing theoretical guarantees of convergence to the optimal policy. When the model set follows a bounded‑parameter MDP structure, the optimization becomes convex, enabling finite‑time convergence under mild assumptions and demonstrating accelerated learning in simulations.
By J. S. van Hulst, W. P. M. H. Heemels, D. J. Antunes
The paper investigates preference-based bandits where a learner selects pairs of arms and receives binary preference feedback modeled by Bradley–Terry. It introduces the locally sensitive eluder dimension, a new complexity measure for logistic preference feedback, and proposes the GINOP algorithm that uses log-loss confidence sets to balance optimism and exploration. The authors prove a first-order regret bound showing that learning with preference feedback can be as statistically efficient as learning from direct rewards, and they validate their theory with empirical experiments.
By Ahmed Ben Yahmed (CREST, ENSAE Paris, FAIRPLAY), Marc Abeille (FAIRPLAY), Cl\'ement Calauz\`enes (FAIRPLAY)
The paper introduces Residual Reward Models (RRM) to enhance preference‑based reinforcement learning (PbRL) in robotics. RRMs decompose the true reward into a prior component—such as a heuristic, language‑generated, or IRL‑derived reward—and a learned residual that is trained with human preferences. Experiments on Meta‑World, DM‑Control, and a physical Franka Panda robot show that RRMs markedly improve sample efficiency and accelerate policy learning compared to standard PbRL methods.
By Chenyang Cao, Miguel Rogel-Garc\'ia, Mohamed Nabail, Xueqian Wang, Nicholas Rhinehart
arXiv:2606. 01382v1 Announce Type: cross Abstract: Preference alignment is central to improving large language models, but standard reward-based formulations can be restrictive when human preferences are cyclic, non-transitive, or otherwise not representable by a scalar reward.
By Tianlong Nan, Xiaopeng Li, Christian Kroer, Tianyi Lin
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:2606. 03962v1 Announce Type: cross Abstract: Classical reinforcement learning (RL) typically seeks a deterministic policy that maximizes the expected sum of a scalar reward.
By Anthony GX-Chen, Ankit Anand, Gheorghe Comanici, Zaheer Abbas, Eser Ayg\"un, David Smalling, Shibl Mourad, Doina Precup, Andr\'e Barreto, Mark Rowland
arXiv:2607. 11432v1 Announce Type: new Abstract: In this work, we study the reinforcement learning (RL) problem from pairwise trajectory comparisons provided by a human expert.
By Simone Drago, Marco Mussi, Leonardo Bianconi, Alberto Maria Metelli