Reward Maximizing

Reward maximization, a core problem in reinforcement learning, aims to design agents that make optimal decisions to achieve the highest cumulative reward in a given environment. Current research focuses on improving learning efficiency, addressing biases in reward models (particularly within large language models), and developing robust algorithms that handle uncertainty and limited data, including approaches incorporating risk assessment and alternative motivational frameworks beyond reward. These advancements are crucial for building reliable and adaptable AI agents for diverse applications, from robotics and game playing to complex decision-making in real-world scenarios.

Papers

November 1, 2024

Unified theory of upper confidence bound policies for bandit problems targeting total reward, maximal reward, and more
Nobuaki Kikkawa, Hiroshi Ohno
Bandit Algorithm High Confidence Prior Policy Bandit Problem Reward Structure Upper Confidence Bound Reward Maximizing

May 28, 2024

An approach to improve agent learning via guaranteeing goal reaching in all episodes
Pavel Osinenko, Grigory Yaremenko, Georgiy Malaniya, Anton Bolychev
Reinforcement Learning Markov Decision Process Constructive Approach Near Optimal Policy Pseudo Goal Uncertain Episode Length Agent Learning Reward Maximizing

May 19, 2024

Large Language Models are Biased Reinforcement Learners
William M. Hayes, Nicolas Yax, Stefano Palminteri
Reinforcement Learning Context Learning Policy Gradient Reward Maximizing

April 7, 2024

Towards Understanding the Influence of Reward Margin on Preference Model Performance
Bowen Qin, Duanyu Feng, Xi Yang
Language Model Reward Model External Influence Reward Estimation Accuracy Reward Maximizing

March 22, 2023

Strategy Synthesis in Markov Decision Processes Under Limited Sampling Access
Christel Baier, Clemens Dubslaff, Patrick Wienhöft, Stefan J. Kiebel
Reinforcement Learning Markov Decision Process Limited Sample Strategy Synthesis Reward Maximizing Interval Markov Decision Process

November 20, 2022

Reward is not Necessary: How to Create a Modular & Compositional Self-Preserving Agent for Life-Long Learning
Thomas J. Ringstrom
Reinforcement Learning Agent System Lifelong Learning Reward Report Modular System Goal Directed Behavior Variational Empowerment Reward Maximizing

November 6, 2022

Wall Street Tree Search: Risk-Aware Planning for Offline Reinforcement Learning
Dan Elbaz, Gal Novik, Oren Salzman
Offline Reinforcement Learning Tree Search Risk Aware Reward Maximizing

Reward Maximizing

Papers

Unified theory of upper confidence bound policies for bandit problems targeting total reward, maximal reward, and more

An approach to improve agent learning via guaranteeing goal reaching in all episodes

Large Language Models are Biased Reinforcement Learners

Towards Understanding the Influence of Reward Margin on Preference Model Performance

Strategy Synthesis in Markov Decision Processes Under Limited Sampling Access

Reward is not Necessary: How to Create a Modular & Compositional Self-Preserving Agent for Life-Long Learning

Wall Street Tree Search: Risk-Aware Planning for Offline Reinforcement Learning