arXiv research

Study optimal reward schemes for inducing desired player performance in risky contests.

The paper proposes a model reward scheme for collaborative ML based on Shapley value and information gain.

Adaptive exploration scheme for evaluating multiple policies with different rewards.

This paper augments the reward received by a reinforcement learning agent with potential functions in order to help the agent learn (possibly stochastic) optimal policies. We show that a potential-based reward shaping scheme is able to preserve optimality of stochastic policies, and demonstrate that the ability of an a…

New algorithm finds optimal policies without knowing reward functions.

Regularized policies are robust to adversarial rewards.

One-bit feedback suffices for a bandit problem's optimal strategy.

New method for LLMs to learn reasoning by optimizing latent variables.

In light of the growing interest in agent-based market models, we bring together several earlier works in which we considered the topic of self-consistent market modelling. Building upon the binary game structure of Challet and Zhang, we discuss generalizations of the strategy reward scheme such that the agents seek to…

New algorithm uses Whittle index to improve Q-learning for restless bandits.

Shaping in humans and animals has been shown to be a powerful tool for learning complex tasks as compared to learning in a randomized fashion. This makes the problem less complex and enables one to solve the easier sub task at hand first. Generating a curriculum for such guided learning involves subjecting the agent to…

SA-GFN corrects biases in GFlowNets due to graph symmetries.

Reward collapse occurs when ranking-based reward models yield uniform rewards for different prompts.

Local search improves GFlowNets' ability to generate high-reward samples.

Paper optimizes recommendation systems for long-term business metrics.

Paper develops robust estimators and strategies for stochastic MABs with heavy-tailed rewards.

New method tracks shifts in infinite-armed bandits without prior knowledge.

Manipulating data, such as weighting data examples or augmenting with new instances, has been increasingly used to improve model training. Previous work has studied various rule- or learning-based approaches designed for specific types of data manipulation. In this work, we propose a new method that supports learning d…

New algorithm for reward-free RL with linear function approximation, reducing sample complexity.

A scheme robust to action erasures improves MAB performance.

Quantum Proof-of-Work uses boson sampling to secure blockchain consensus.

Top-k Combinatorial Bandits generalize multi-armed bandits, where at each round any subset of $k$ out of $n$ arms may be chosen and the sum of the rewards is gained. We address the full-bandit feedback, in which the agent observes only the sum of rewards, in contrast to the semi-bandit feedback, in which the agent obse…

A novel approach uses an ensemble of Gaussian processes for robust and adaptive reinforcement learning.

We introduce a method by which a generative model learning the joint distribution between actions and future states can be used to automatically infer a control scheme for any desired reward function, which may be altered on the fly without retraining the model. In this method, the problem of action selection is reduce…

Efficient exploration improves large language model performance with fewer queries.

New algorithm for multi-player bandits with collision-dependent rewards.

New method uses simple sensor intentions to learn complex tasks.

Novel approach to contextual bandits using self-supervised learning.

Study on private and robust multi-armed bandits with contaminated heavy-tailed rewards.

We consider the task of Inverse Reinforcement Learning in Contextual Markov Decision Processes (MDPs). In this setting, contexts, which define the reward and transition kernel, are sampled from a distribution. In addition, although the reward is a function of the context, it is not provided to the agent. Instead, the a…

A bandit algorithm reduces regret in noisy, communication-constrained feedback.

A new insurance and reinsurance pricing scheme based on realized loss.

RL framework optimizes trading costs in noisy markets.

Recent reinforcement learning algorithms, though achieving impressive results in various fields, suffer from brittle training effects such as regression in results and high sensitivity to initialization and parameters. We claim that some of the brittleness stems from variance differences, i.e. when different environmen…

In this paper, we present a discrete-type approximation scheme to solve continuous-time optimal stopping problems based on fully non-Markovian continuous processes adapted to the Brownian motion filtration. The approximations satisfy suitable variational inequalities which allow us to construct $ε$-optimal stopping tim…

New method for efficient online exploration in RLHF reduces regret.

New algorithm for collaborative bandit learning reduces sample complexity and regret.

Interactive RL and DT feedback improve feature selection efficiency.

Paper closes the gap in MP-MAB problems with novel adaptive communication and exploration.

In this paper, a new population-guided parallel learning scheme is proposed to enhance the performance of off-policy reinforcement learning (RL). In the proposed scheme, multiple identical learners with their own value-functions and policies share a common experience replay buffer, and search a good policy in collabora…

This paper presents four different ways of looking at the well-known Least Squares Temporal Differences (LSTD) algorithm for computing the value function of a Markov Reward Process, each of them leading to different insights: the operator-theory approach via the Galerkin method, the statistical approach via instrumenta…

Improved DPO framework penalizes preference uncertainty to avoid overoptimization.

Game theory shows miners' hardware improvements don't centralize mining.

This work optimizes identifying good arms in nonparametric multi-armed bandits.

Efficiently learns matching rewards in two-sided markets with matrix completion.

Algorithm learns diffusion processes with high-dimensional state spaces.

This paper optimizes driver repositioning using MARL and reward design for better service and traffic management.

New algorithm LOOP learns infinite-horizon AMDPs efficiently with function approximation.