HyperX uses reward bonuses to enable efficient exploration in meta-learning.
problem Catastrophic failure of meta-learning with sparse rewards.
method HyperX uses novel reward bonuses to explore in approximate hyper-state space.
result HyperX meta-learns better task-exploration and adapts more successfully to new tasks.
New exploration bonuses improve reinforcement learning efficiency.
problem Efficient exploration in unknown environments with limited feedback.
method Improved exploration bonuses scaling with 1/n and improved stopping time analysis.
result Faster learning rates and improved sample complexity in pure-exploration settings.
The paper evaluates various bonus-based exploration methods in the ALE and finds limited improvement in performance.
problem Improving exploration in reinforcement learning algorithms, especially in challenging games.
method Empirical evaluation of different reward bonuses on the Arcade Learning Environment.
result Recently developed bonus-based exploration methods do not significantly improve performance in challenging games.
Directed exploration improves reinforcement learning efficiency and robustness.
problem Achieving good sample efficiency in reinforcement learning with efficient exploration.
method Directed exploration through goal-conditioned policies that are independent of uncertainty.
result Directed exploration is more efficient and robust to uncertainty than reward bonuses.
Paper simplifies complex AI exploration by predicting future rewards.
problem Training machines to optimally gather complex information.
method Developed a denser reward structure using cross-value to decouple exploration and exploitation.
result Demonstrated successful learning of challenging tasks without shaping or bonuses.
Achieving efficient and scalable exploration in complex domains poses a major challenge in reinforcement learning. While Bayesian and PAC-MDP approaches to the exploration problem offer strong formal guarantees, they are often impractical in higher dimensions due to their reliance on enumerating the state-action space.…
Paper proposes a method for learning and planning in time-varying environments.
problem Learning and planning in unknown, time-varying environments.
method Computes the maximally likely model of the environment using maximum likelihood estimation.
result Generalizes learning algorithms for time-invariant Markov decision processes to time-varying ones.
MADE improves exploration in RL by maximizing deviation from explored regions.
problem Efficient exploration in high-dimensional RL tasks with sparse rewards.
method Proposes a new exploration approach via maximizing the deviation of the occupancy of the next policy from explored regions, adding it as an adaptive regularizer to the RL objective.
result Significantly improves sample efficiency in navigation and locomotion tasks.
Proposes Optimistic Pessimistically Initialised Q-Learning (OPIQ) for better exploration in RL.
problem Pessimistic initialisation of Q-values in deep RL leads to poor exploration performance.
method Augments pessimistically initialised Q-values with count-based bonuses to ensure optimism.
result OPIQ outperforms non-optimistic DQN variants in hard exploration tasks.
Improved exploration in adversarial MDPs via dilated bonuses.
problem Global exploration in adversarial MDPs with local-search policies.
method Dilated bonuses to facilitate global exploration.
result Improved regret bounds in various MDP settings.
All reinforcement learning algorithms must handle the trade-off between exploration and exploitation. Many state-of-the-art deep reinforcement learning methods use noise in the action selection, such as Gaussian noise in policy gradient methods or ε ε ε -greedy in Q-learning. While these methods are appealing due to their…
This paper improves Q-learning bounds using reference-advantage decomposition.
problem Improving Q-learning bounds in MDPs with positive suboptimality gaps.
method Develops a novel error decomposition framework to prove gap-dependent regret bounds.
result Establishes logarithmic gap-dependent regret bounds for Q-learning.
Most of the prior work on multi-agent reinforcement learning (MARL) achieves optimal collaboration by directly controlling the agents to maximize a common reward. In this paper, we aim to address this from a different angle. In particular, we consider scenarios where there are self-interested agents (i.e., worker agent…
RP1 uses active learning to improve world model in fewest samples.
problem Improving sample efficiency in MBRL for continuous control tasks.
method RP1 views MBRL as active learning, using a hybrid objective function and principled termination.
result Statistically significant gains over existing approaches on continuous control tasks.
New algorithm optimizes exploration in unknown MDPs.
problem Optimal exploration in unknown Markov decision processes.
method Trajectory-optimization-based approximate solution using Bayesian experimental design.
result Empirically superior to intrinsic-motivation-based algorithms in convergence and model fidelity.
New approach incentivizes strategic agents to explore, making exploration almost free.
problem Incentivized exploration in multi-armed bandits with long-term strategic agents.
method Simple incentive-provision strategy, best arm identification algorithm, and UCB lower bound.
result Exploration can be (almost) free when there are many learning agents.
Improved privacy in RL with near-optimal regret bounds.
problem Privacy-preserving reinforcement learning in personalized decision-making systems.
method Differentially private algorithm based on LSVI-UCB++ with privacy-preserving techniques.
result Achieved a near-optimal regret bound of O(d * sqrt(H^3 * K) + H^(15/4) * d^(7/6) * K^(1/2) / ε).
New algorithms achieve logarithmic regret in KL-regularized Markov games.
problem Improving sample efficiency in game-theoretic settings with KL regularization.
method Developed OMG and SOMG algorithms for matrix and Markov games, using best response sampling and superoptimistic bonuses.
result Logarithmic regret in T T T that scales inversely with KL regularization strength β β β . New algorithm tackles multi-agent reinforcement learning with optimal convergence rate.
problem Multi-agent reinforcement learning with large state spaces and linear function approximations.
method Refined AVLPR framework with data-dependent pessimistic estimation and action-dependent bonuses.
result First algorithm with optimal O ( T − 1 / 2 ) O(T^{-1/2}) O ( T − 1/2 ) convergence rate and no poly( A max A_{\max} A m a x ) dependency. New methods tackle robust reinforcement learning in sparse, corrupted data.
problem Tackles robust reinforcement learning in sparse, corrupted data.
method Proposes actor-critic methods with sparse robust estimator oracles.
result First non-vacuous guarantees in high-dimensional sparse MDPs with single-policy concentrability coverage.
Langevin DQN achieves deep exploration using Gaussian noise.
problem Deep exploration in reinforcement learning.
method Developed Langevin DQN, a variation of DQN with Gaussian noise.
result Langevin DQN achieves deep exploration.
In this paper, we focus on a prediction-based novelty estimation strategy upon the deep reinforcement learning (DRL) framework, and present a flow-based intrinsic curiosity module (FICM) to exploit the prediction errors from optical flow estimation as exploration bonuses. We propose the concept of leveraging motion fea…
We consider the exploration/exploitation problem in reinforcement learning. For exploitation, it is well known that the Bellman equation connects the value at any time-step to the expected value at subsequent time-steps. In this paper we consider a similar \textit{uncertainty} Bellman equation (UBE), which connects the…
Improved reinforcement learning algorithm with linear approximation for unknown dynamics.
problem Reinforcement learning with adversarial changing cost functions and bandit feedback.
method Combines mirror-descent and least squares policy evaluation in an auxiliary MDP.
result Obtains an O ~ ( K 6 / 7 ) \widetilde O(K^{6/7}) O ( K 6/7 ) regret bound, significantly improving over previous methods. The paper analyzes the intrinsic exploration terms in policy-gradient algorithms.
problem Exploration in policy-gradient algorithms and its impact on policy optimization.
method Numerical optimization criteria and stochastic gradient analysis.
result Exploration techniques improve policy optimization by smoothing the learning objective and modifying gradient estimates.
Reward hacking exploits misspecified rewards, affecting agent capabilities and true performance.
problem Reward hacking in RL models exploiting reward misspecifications.
method Constructed four RL environments with misspecified rewards; analyzed agent capabilities and behavior.
result More capable agents exploit reward misspecifications, achieving higher proxy reward but lower true reward.
New method improves RL in continuous spaces with kernel smoothing.
problem Sample efficiency and structural assumptions in classical RL.
method Kernel smoothing model-based approach with Bernstein-style exploration bonus.
result Achieves improved regret bound in finite-horizon settings.
Paper introduces PRMs to learn non-Markovian stochastic rewards for reinforcement learning.
problem Lack of structured representation for non-Markovian stochastic rewards in reinforcement learning.
method Introduces probabilistic reward machines (PRMs) and presents an algorithm to learn them from decision processes.
result Algorithm proves correct and convergent for learning PRMs from decision processes.
Paper addresses reward learning issues in RL, improving both under- and over-estimation.
problem Reward learning from data can lead to reward delusions or underestimation, causing unintended behaviors.
method Connects reward learning to positive-unlabeled (PU) learning and applies a large-scale PU learning algorithm.
result Improves both GAIL and supervised reward learning without additional assumptions.
This work analyzes the value of future reward information in RL.
problem Analyzing the impact of knowing future rewards in reinforcement learning.
method Competitive analysis and worst-case reward distribution.
result Exact ratios between standard RL agents and those with future-reward lookahead.
Self-supervised reward prediction improves RL in sparse reward settings.
problem Data efficiency and sparse reward signals in reinforcement learning.
method Learning a state representation for reward prediction and using it to shape rewards.
result Self-supervised reward prediction enhances RL algorithms in single-goal environments.
The study categorizes reward errors in reinforcement learning, finding some can be beneficial.
problem Training language models with imperfect proxy rewards.
method Theoretical analysis of policy gradient optimization and categorization of reward errors.
result Reward errors can be benign or even beneficial, preventing policy from stalling.
Reward collapse occurs when ranking-based reward models yield uniform rewards for different prompts.
problem Reward collapse in aligning large language models with human preferences.
method Introduced a prompt-aware optimization scheme to derive closed-form expressions for reward distributions.
result Our prompt-aware utility functions significantly alleviate reward collapse during training.
Reward models need more than just accuracy for effective RLHF.
problem The effectiveness of reward models in RLHF is not fully understood.
method An optimization perspective to evaluate reward models.
result Reward models with low reward variance can lead to a flat optimization landscape, hindering performance.
Proposes a method to boost deep reinforcement learning with sparse rewards.
problem Challenges in learning complex behaviors with long horizons and sparse rewards.
method Predictive coding for reward shaping.
result Achieves better learning by providing reward signals that understand environment dynamics and emphasize useful features.
Action guidance helps agents learn true objectives in games with sparse rewards.
problem Training agents in games with sparse rewards requires significant exploration.
method Action guidance, a novel technique that combines exploration with reward shaping.
result Action guidance enables agents to optimize true objectives efficiently.
New RL method uses distance between states instead of rewards for sparse reward environments.
problem Sparse rewards or non-reward environments in reinforcement learning.
method Uses goal-distance gradient and bridge point planning for policy improvement.
result Significantly better performance on sparse reward and local optimal problems in complex environments.
Paper proposes RRD to learn proxy rewards for sparse delayed rewards in episodic reinforcement learning.
problem Learning from sparse and delayed rewards in reinforcement learning.
method Randomized Return Decomposition (RRD) algorithm to redistribute rewards.
result Substantial improvement over baseline algorithms in experiments.
Enhances reward specification in RL with a novel language-based approach.
problem Reward specification in RL can lead to unintended, potentially harmful behaviours.
method Developed a novel class of language-based Reward Machines using RML's built-in memory.
result Can specify non-regular, non-Markovian reward functions for complex tasks.
Reward tweaking optimizes behavior for long-term goals by adjusting the reward function.
problem Optimizing behavior for long-term goals in reinforcement learning with unstable long planning horizons.
method Reward tweaking learns a surrogate reward function that induces optimal behavior for the original task.
result Reward tweaking guides agents towards better long-term returns while planning for short horizons.
We propose a generic, Bayesian, information geometric approach to the exploration--exploitation trade-off in multi-armed bandit problems. Our approach, BelMan, uniformly supports pure exploration, exploration--exploitation, and two-phase bandit problems. The knowledge on bandit arms and their reward distributions is su…
Many reinforcement-learning researchers treat the reward function as a part of the environment, meaning that the agent can only know the reward of a state if it encounters that state in a trial run. However, we argue that this is an unnecessary limitation and instead, the reward function should be provided to the learn…
Extends reinforcement learning alignment to scalar rewards, improving math reasoning.
problem Designing reinforcement learning algorithms for general LLM alignment.
method Introduces f-GRPO and f-HAL, estimating f-divergences between reward-aligned and unaligned distributions.
result Improves math reasoning RLVR tasks and mitigates reward hacking.
Study differential privacy in multi-agent RL, achieving efficient and private learning.
problem Protecting sensitive data in multi-agent reinforcement learning.
method Extending DP definitions to two-player games, designing an efficient algorithm with privatized bonuses.
result Achieved trajectory-wise differential privacy in multi-agent RL, improving regret bounds.
This work characterizes reward function partial identifiability and its impact on policy optimization.
problem Reward function partial identifiability in complex tasks.
method Formal characterisation of partial identifiability using various reward learning data sources.
result Unified framework for comparing data sources and downstream tasks by their invariances.
This paper introduces a new reward shaping method for average-reward reinforcement learning.
problem Speeding up convergence to an optimal policy in average-reward reinforcement learning tasks.
method Developed a temporal logic-based approach to automatically generate reward shaping functions.
result The optimal policy can be recovered using the proposed reward shaping framework.
We consider the problem of provably optimal exploration in reinforcement learning for finite horizon MDPs. We show that an optimistic modification to value iteration achieves a regret bound of O ~ ( H S A T + H 2 S 2 A + H T ) \tilde{O}( \sqrt{HSAT} + H^2S^2A+H\sqrt{T}) O ~ ( H S A T + H 2 S 2 A + H T ) where H H H is the time horizon, S S S the number of states, A A A the number of action…
Paper generalizes reward distribution in multi-armed bandits with temporally-partitioned rewards.
problem Handling partial rewards distributed over multiple rounds in multi-armed bandits.
method Introduces Beta-spread property to generalize reward distribution, derives lower bound, and provides TP-UCB-FR-G algorithm.
result Improves regret upper bound for some scenarios using Beta-spread property.