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…
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
168,932 papers · 148 categories
Trend · papers per month
48 results for “Projection Rewards”
The paper tackles linear bandits with projections, achieving optimal regret.
problem Learning the best arm in a linear bandit model with unobservable projection rewards.
method Developed strategies for both finite and infinite arms, achieving optimal regret bounds.
result Achieved optimal regret bounds for both finite and infinite arms.
Project explores reinforcement learning solutions for sparse reward environments.
problem Difficulty in navigating environments with infrequent rewards.
method Contrast and investigate existing reinforcement learning solutions in various video games.
result Introduces a novel reinforcement learning solution combining curiosity and auxiliary tasks.
DDO-RM improves reward-based policies by converting reward scores into a target distribution.
problem Improving reward-based policies when the reward function is simpler than the policy.
method Converts reward scores into a target distribution and uses KL-regularized mirror-descent updates.
result DDO-RM outperforms DPO in pair accuracy and mean margin.
New algorithms for multivariate RL improve decision-making in complex systems.
problem Complex multi-objective decision-making in reinforcement learning.
method Oracle-free and computationally-tractable algorithms for multivariate distributional RL.
result Convergence rates match scalar reward settings and provide insights into reward dimensionality.
New algorithm reduces regret for single-index bandits to nearly optimal.
problem Optimizing rewards from unknown projections of high-dimensional contexts.
method Two-phase algorithm: estimate projection direction, reduce to 1D bandit, use UCB.
result Achieved optimal regret of .
Improved regret for single-index bandits with optimal algorithm.
problem Optimizing rewards from unknown one-dimensional projections of high-dimensional contexts.
method Two-phase algorithm: first estimate projection direction, then discretize and use UCB.
result Achieved optimal regret of .
This paper studies GAIL's global convergence for general MDP and nonlinear rewards.
problem Understanding when GAIL algorithms achieve global convergence for general MDP and nonlinear rewards.
method Characterization of global convergence for various policy gradient algorithms applied to GAIL.
result First systematic theoretical study of GAIL for global convergence.
New algorithm achieves optimal regret in average reward MDPs without prior bias information.
problem Achieving optimal regret in average reward MDPs with computational efficiency and without prior bias information.
method Projective Mitigated Extended Value Iteration (PMEVI) to compute bias-constrained optimal policies efficiently.
result First tractable algorithm with minimax optimal regret of .
CausalRM models rewards from user feedback, overcoming noise and bias.
problem Aligning language models with user preferences from noisy, biased feedback.
method Causal-theoretic reward modeling framework addressing noise and bias in observational feedback.
result CausalRM learns accurate reward signals from noisy and biased observational feedback.
New algorithm for differential private kernel bandits with improved regret bounds.
problem Contextual bandits with differential privacy constraints.
method Proposes a private kernel-ridge regression estimator using random projections.
result Achieves state-of-the-art cumulative regret bounds in both joint and local models of differential privacy.
Algorithm optimizes system design and control for better rewards.
problem Optimizing system design and control for maximum rewards.
method Deep reinforcement learning combining policy gradient and model-based optimization.
result DEPS algorithm outperforms state-of-the-art methods in various environments.
Small Total-Cost Constraints in Contextual Bandits with Knapsacks, with Application to Fairnessstat.ML
New method reduces total cost constraints in CBwK to sqrt(T) with fairness application.
problem Maximize rewards while adhering to total cost constraints in CBwK.
method Dual strategy based on projected-gradient-descent updates.
result Total cost constraints reduced to sqrt(T) with poly-logarithmic terms.
Improves inference-time alignment for diffusion models without updating weights.
problem Aligning diffusion models without updating weights for high-reward outputs.
method Trust-Region Iterative Twisted Sequential Monte Carlo (TRI-TSMC) for variance reduction and efficiency.
result Improves primary alignment objectives on text generation tasks.
Automatically learns dynamical distances for efficient reinforcement learning.
problem Difficult to specify reward functions for reinforcement learning.
method Uses dynamical distances to provide well-shaped reward functions.
result Can learn complex tasks efficiently with minimal supervision.
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…
Zero-shot Policy Learning with Spatial Temporal RewardDecomposition on Contingency-aware Observationcs.LG
Paper tackles zero-shot policy learning for agents in new environments.
problem Learning in one environment and generalizing to unseen environments without additional data.
method Decomposes sparse rewards into spatial-temporal regions and uses Model Predictive Control (MPC).
result Method achieves natural generalization to new environments, demonstrated on various tasks.
FORE evaluates occupancy ratios without requiring Bellman completeness.
problem Offline reinforcement learning occupancy ratio estimation.
method Fitted occupancy-ratio evaluation (FORE) using adjoint Bellman recursion.
result FORE achieves convergence in KL without Bellman completeness.
New algorithm for nonstationary GLBs reduces computation and memory costs.
problem Nonstationary generalized linear bandits with unknown time-varying parameters.
method Discounted Online Mirror Descent (DOMD) for parameter estimation.
result Dynamic regret bounds of order per round in drifting and piecewise-stationary environments.
Study intrinsic motivation for synergistic tasks in reinforcement learning.
problem Sparse-reward synergistic tasks where multiple agents must work together.
method Propose incentivizing actions that affect the world in ways not achievable individually, using either true states or a dynamics model.
result Our approach yields more efficient learning than typical methods.
Nonparametric Thompson Sampling achieves optimal regret for risk-averse bandits with sub-Gaussian rewards.
problem Optimizing risk-averse bandit problems with sub-Gaussian rewards.
method Anchor-free nonparametric Thompson Sampling algorithm .
result Achieves regret matching the instance-dependent lower bound to leading order in .
Kernel method estimates long-term effects from short-term data.
problem Estimating long-term effects from short-term data in continuous actions.
method Kernel ridge regression to embed and extrapolate long-term effects.
result Uniform consistency and nonasymptotic error bounds for the estimator.
TROLL improves RL for LLMs by replacing clipping with a trust region projection.
problem Clipping in RL for LLMs causes instability and suboptimal performance.
method TROLL uses a discrete differentiable trust region projection to replace clipping, balancing computational cost and effectiveness.
result TROLL consistently outperforms PPO-like clipping in training speed, stability, and final success rates.
Thresholded Lasso Banditstat.ML
Thresholded Lasso bandit minimizes regret in sparse linear bandits.
problem Sparse stochastic contextual linear bandits with large feature vectors.
method Uses Lasso framework with thresholding to estimate reward function and its sparse support.
result Non-asymptotic regret upper bounds scaling as .
Kernel-based methods improve policy evaluation in MRP models.
problem Estimating value functions in infinite-horizon discounted MRP models.
method Kernel-based temporal difference methods using reproducing kernel Hilbert spaces.
result Optimal error bounds derived for the kernel-based LSTD estimate.
A stochastic model helps maintain insufficiently funded pension funds.
problem Maintaining pension funds that are underfunded and require external financing.
method A time-homogeneous diffusion process with a barrier is used to model the unrestricted reserves value, and a renewal-reward process models the financing effort.
result Expected values and cost evaluations of maintenance are derived, and the approach is applied to a generalized Brownian motion process.
New method decomposes Markov chain rewards into persistent and transient components.
problem Ambiguity in classical evaluation methods for Markov chains with reducible and periodic states.
method Minimal exact quotient by the real peripheral invariant subspace, decomposing rewards into persistent and transient components.
result Exact comparison with classical methods shows that the new decomposition reallocates the same information, making persistent modes explicit.
New algorithm achieves asymptotically optimal regret without horizon dependence.
problem Horizon-free regret minimization for reinforcement learning.
method Proposes a new algorithm and proves a regret upper bound.
result Regret upper bound of \(\tilde O(\sqrt{SAK} + S^8A^3)\) with failure probability \(\delta\).
This project combines recent advances in experience replay techniques, namely, Combined Experience Replay (CER), Prioritized Experience Replay (PER), and Hindsight Experience Replay (HER). We show the results of combinations of these techniques with DDPG and DQN methods. CER always adds the most recent experience to th…
Despite the success of single-agent reinforcement learning, multi-agent reinforcement learning (MARL) remains challenging due to complex interactions between agents. Motivated by decentralized applications such as sensor networks, swarm robotics, and power grids, we study policy evaluation in MARL, where agents with jo…
FOCOPS optimizes agent's behavior while adhering to constraints.
problem Optimizing agent's behavior while respecting safety constraints.
method FOCOPS solves a constrained optimization problem in policy space, then projects the solution back into the parametric space.
result FOCOPS achieves better performance on constrained robotics tasks.
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.
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: Getting the Best of Sparse Rewards and Shaped Rewards for Real-time Strategy Gamescs.LG
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.
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…
f-GRPO and Beyond: Divergence-Based Reinforcement Learning Algorithms for General LLM Alignmentcs.LG
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.
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.