StepGAN improves conditional sequence generation with less computation.
problem Stabilizing training of sequence generative adversarial networks (SeqGAN).
method Modified discriminator assigns scores at each generation step.
result StepGAN outperforms previous methods on synthetic and chit-chat tasks.
Reward-poisoning attacks can force RL agents to learn bad policies, and we categorize and quantify their feasibility.
problem Reward-poisoning attacks can manipulate RL agents to learn undesirable policies.
method Categorize attacks by infinity-norm constraint, provide thresholds for feasibility, and develop adaptive attack strategies.
result Adaptive reward-poisoning attacks can achieve the nefarious policy in polynomial steps, while non-adaptive attacks require exponential steps.
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.
A multi-player bandit system resists adversarial attacks with near-optimal regret.
problem Adversaries attempt to manipulate rewards in a multi-player multi-armed bandit game.
method Players communicate a single bit to resist attacks, achieving near-optimal regret.
result Achieves near-optimal regret of O ( log 1 + δ T + W ) O(\log^{1+δ}T + W) O ( log 1 + δ T + W ) , where W W W is the total time of adversarial attacks. A novel framework refines diffusion models iteratively for better downstream reward optimization.
problem Optimizing reward functions during inference of diffusion models.
method Iterative refinement process with noising and reward-guided denoising steps.
result Superior empirical performance in protein and DNA design.
New TD method stabilizes average-reward learning.
problem Stability issues in average-reward TD learning.
method Implicit fixed point update for average-reward TD( λ λ λ ). result Improved numerical stability and broader step-size range.
We consider the Max K K K -Armed Bandit problem, where a learning agent is faced with several sources (arms) of items (rewards), and interested in finding the best item overall. At each time step the agent chooses an arm, and obtains a random real valued reward. The rewards of each arm are assumed to be i.i.d., with an un…
Algorithm learns near-optimal policies for reward-mixing MDPs with few latent contexts.
problem Episodic reinforcement learning in reward-mixing Markov decision processes with a few latent contexts.
method Sample-efficient algorithm EM^2 using higher-order method-of-moments approach.
result Provides an ε-optimal policy using O(ε^(-2) * S^d A^d * poly(H, Z)^d) episodes for arbitrary M ≥ 2.
New algorithm learns reward signals from episodic returns for better reinforcement learning.
problem Difficulty in designing reward functions for real-world reinforcement learning tasks.
method Introduces a new algorithm that decomposes episodic returns into time-step rewards using deep neural networks.
result Learning reward signals from episodic returns improves reinforcement learning efficiency.
MFMs enable efficient reward alignment for generative models.
problem Computational bottleneck in controlling generative models.
method Meta Flow Maps (MFMs) extend consistency models and flow maps to stochastic regime for efficient value function estimation.
result MFMs enable inference-time steering and unbiased, off-policy fine-tuning to general rewards efficiently.
Demon aligns diffusion models without retraining or backpropagation.
problem Aligning diffusion models with user preferences.
method Stochastic optimization to control noise distribution.
result Significantly improves aesthetics scores for text-to-image generation.
New algorithms improve efficiency in learning from personalized rewards.
problem Learning from personalized rewards in recommendation systems.
method Developed provably efficient algorithms with sublinear regret for context-dependent feedback.
result Introduced a Lipschitz reward estimator that improves generalization performance.
Align-RUDDER improves reinforcement learning with few demonstrations by redistributing rewards.
problem Learning complex tasks with sparse and delayed rewards using few demonstrations.
method Align-RUDDER uses a profile model for reward redistribution based on multiple sequence alignment of demonstrations.
result Align-RUDDER outperforms competitors on complex tasks with few demonstrations.
A new estimator for state values in reinforcement learning reduces complexity and improves convergence.
problem Estimating state values in reinforcement learning with Markov reward processes.
method Loop estimator exploiting regenerative structure of Markov reward processes.
result Instance-dependent convergence rate of O ~ ( τ s / T ) \widetilde{O}\left(\sqrt{τ_s/T}\right) O ( τ s / T ) for estimating state values. Improved RL training for DMs reduces mode collapse and preserves diversity.
problem Mode collapse and training instability in RL fine-tuned diffusion models.
method Dynamic hierarchical RL training with sliding-window parameter regularisation.
result Models trained with HRF achieve better preservation of diversity in downstream tasks.
The paper analyzes RLVR's training dynamics, proving convergence depends on aligning update direction with Gradient Gap.
problem Understanding why RLVR works and its limitations.
method Analysis of RLVR's training process at trajectory and token levels, introducing Gradient Gap.
result Convergence depends on aligning update direction with Gradient Gap, with a sharp step-size threshold.
The paper shows how curated synthetic data can optimize human preferences in generative models.
problem Contamination of web-scale datasets by synthetic data affects future model training.
method Theoretical study of iterated retraining of generative models with curated synthetic data.
result Data curation can be seen as an implicit preference optimization mechanism, maximizing expected reward.
New algorithm for non-stationary bandits with slow drifts.
problem Minimizing dynamic regret in non-stationary bandits with slowly varying rewards.
method Extends Successive Elimination to non-stationary bandits with a novel gap profile characterization.
result First instance-dependent regret upper bound for slowly varying non-stationary bandits.
A new algorithm balances exploration and exploitation in online decision-making.
problem Balancing exploration and exploitation in online decision-making.
method Proposed C 4 C^4 C 4 -UCB algorithm incorporating conservative mechanism. result Proved n-step upper regret bound for two situations.
BatchGFN uses generative flow networks for efficient batch active learning.
problem Efficiently selecting informative batches for active learning.
method Generative flow networks to sample batches proportional to a batch reward.
result Constructs highly informative batches with a single forward pass per point.
New RLHF algorithm identifies optimal policies from human feedback without explicit reward inference.
problem Training large language models with human feedback without reward inference.
method Model-free RLHF algorithm B S A D \mathsf{BSAD} BSAD that identifies optimal policies directly from human preference. result Provable, instance-dependent sample complexity i l d e O ( c M S A 3 H 3 M log 1 δ ) ilde{\mathcal{O}}(c_{\mathcal{M}}SA^3H^3M\log\frac{1}δ) i l d e O ( c M S A 3 H 3 M log δ 1 ) . Novel alignment framework for text-to-image generation using diffusion models and flow matching.
problem Improving text-to-image generation with minimal computational resources.
method Proposes a novel alignment framework that decomposes the score function into pre-trained score plus a conditional expectation of the reward.
result Achieves comparable performance to finetuning-based models with reduced computational cost.
Paper presents a modular RL framework for Forex trading, addressing limitations of prior studies.
problem Challenges in applying RL to Forex trading, including unrealistic environments, simplified rewards, and restricted action spaces.
method Integrates three components: a friction-aware execution engine, a decomposable reward architecture, and a discrete action interface.
result Empirical evaluation shows strong non-monotonic reward interactions and optimal Sharpe ratio with the full reward configuration.
We introduce two tactics to attack agents trained by deep reinforcement learning algorithms using adversarial examples, namely the strategically-timed attack and the enchanting attack. In the strategically-timed attack, the adversary aims at minimizing the agent's reward by only attacking the agent at a small subset of…
ETGL-DDPG improves DDPG for sparse reward control with new exploration and replay techniques.
problem Sparse reward continuous control in reinforcement learning.
method Introduces ε t εt εt -greedy search and GDRB framework for efficient exploration and reward use. result ETGL-DDPG outperforms DDPG and other methods on sparse-reward continuous benchmarks.
Multi-armed bandit (MAB) is a class of online learning problems where a learning agent aims to maximize its expected cumulative reward while repeatedly selecting to pull arms with unknown reward distributions. We consider a scenario where the reward distributions may change in a piecewise-stationary fashion at unknown …
EPIC quantifies reward differences without policy optimization.
problem Distinguishing reward function quality from policy optimization issues.
method EPIC distance to compare reward functions directly.
result EPIC bounds policy training success and regret.
New method learns agent actions from raw video demonstrations.
problem Imitation learning for natural-looking movements.
method Generative Adversarial Networks (GANs) for joint state and reward learning.
result Adversarial imitation learning on raw videos produces similar performance to state-of-the-art methods.
We consider the Max K K K -Armed Bandit problem, where a learning agent is faced with several stochastic arms, each a source of i.i.d. rewards of unknown distribution. At each time step the agent chooses an arm, and observes the reward of the obtained sample. Each sample is considered here as a separate item with the rewa…
The Multi-Armed Bandits (MAB) framework highlights the tension between acquiring new knowledge (Exploration) and leveraging available knowledge (Exploitation). In the classical MAB problem, a decision maker must choose an arm at each time step, upon which she receives a reward. The decision maker's objective is to maxi…
UCRL-CMDP algorithm optimizes RL with constraints on average costs.
problem Optimizing RL in MDPs with average cost constraints.
method Model-based RL algorithms maximizing reward while keeping costs within bounds.
result UCRL-CMDP algorithm's expected regret is upper-bounded by $T^{2\slash 3}$ .
New MAB problem with delayed, anonymous feedback analyzed.
problem Delayed, anonymous feedback in stochastic bandits.
method Phase-based extensions of UCB algorithm for SDCAF.
result Sub-linear regret guarantees for proposed algorithms.
Develops neural network framework for risk-reward optimization problems.
problem Multi-period risk-reward optimization with constrained policies.
method Neural network framework with two coupled feedforward networks, parametrizing two-step policies.
result Empirical optimum converges to true optimal value as network capacity and training size increase.
New RL method MAC improves performance in sparse reward settings.
problem Slow mixing in large state spaces or sparse rewards.
method Multi-level Monte Carlo Actor-Critic (MAC) algorithm.
result Achieves convergence rate comparable to state-of-the-art AC algorithms.
Proposes Constrained Q-learning for reinforcement learning with constraints.
problem Optimizing multiple objectives while adhering to constraints in reinforcement learning.
method Directly restricts the action space in Q-update to learn optimal Q-function for constrained MDP.
result Improves safety and optimality in high-level decision making for autonomous driving.
Improved multi-step TD learning with control variates reduces variance and improves performance.
problem Variance in multi-step TD learning causes divergence in off-policy settings.
method Per-decision control variates for multi-step TD algorithms.
result Control variates significantly improve performance in both on and off-policy tasks.
Novel algorithm reduces computational burden in IRL with finite-time guarantees.
problem Efficiently recover reward function and optimal policy from expert behavior.
method Single-loop algorithm that maximizes likelihood after each policy improvement step.
result Algorithm provably converges to a stationary solution with finite-time guarantees.
A new EM framework for goal-conditioned RL improves performance on sparse reward tasks.
problem Handling sparse rewards in goal-conditioned reinforcement learning.
method A graphical model framework with an EM algorithm that includes a learning-in-hindsight E-step and a supervised M-step.
result hEM significantly outperforms model-free baselines on goal-conditioned benchmarks with sparse rewards.
A novel bandit problem with context-dependent rewards and blocking.
problem Contextual bandit problem with blocking.
method Online bipartite matching, UCB algorithm, delayed exploitation, opportunistic subsampling.
result Guaranteed O ( log T ) \mathcal{O}(\log T) O ( log T ) -regret in bandit setting. Robust Q-learning algorithm resists corrupted rewards.
problem Non-ideal environments with corrupted rewards.
method Developed a novel robust Q-learning algorithm using historical data.
result Finite-time convergence rate matches state-of-the-art bounds.
Deep reinforcement learning improves classification accuracy for imbalanced datasets.
problem Imbalanced datasets challenge conventional classification algorithms.
method Formulated as a sequential decision-making process, solved using deep Q-learning network.
result Proposed model outperforms other imbalanced classification algorithms.
Optimal best arm identification for multi-objective bandits with fixed error probability.
problem Identifying the best arm for each of multiple objectives with fixed confidence.
method Surrogate proportions to sample arms at each time step, eliminating max-min optimisation.
result Asymptotically optimal algorithm for multi-objective best arm identification.
A new reinforcement learning framework for multi-reward processing.
problem Understanding and modeling multi-reward interactions in complex systems.
method Two-stream reward processing with biological associations.
result Agents can react differently to different types of rewards.
A UCB algorithm reduces regret in cooperative multi-agent graph bandits.
problem Cooperative multi-agent decision-making on a graph with shared rewards.
method Upper Confidence Bound (UCB) algorithm for minimizing regret.
result The Multi-G-UCB algorithm achieves expected regret O ( γ N log ( T ) [ K T + D K ] ) O(γN\log(T)[\sqrt{KT} + DK]) O ( γ N log ( T ) [ K T + D K ]) . Algorithm estimates human decision-making in high-dimensional states with finite-time guarantees.
problem Estimating optimal policies and measures of fit in dynamic decision models with high-dimensional state spaces.
method Single-loop estimation algorithm with stochastic gradient steps for reward maximization.
result Algorithm converges to a stationary solution with finite-time guarantees and approximates maximum likelihood sublinearly.
New algorithm for combinatorial bandit problems reduces regret.
problem Optimal selection of sets of arms in bandit problems.
method SGB algorithm with optimized exploration of unselected arms.
result Achieves ( 1 − 1 / e ) (1-1/e) ( 1 − 1/ e ) -regret bound of O ( n 1 3 k 2 3 T 2 3 log ( T ) 2 3 ) \mathcal{O}(n^{\frac{1}{3}} k^{\frac{2}{3}} T^{\frac{2}{3}} \log(T)^{\frac{2}{3}}) O ( n 3 1 k 3 2 T 3 2 log ( T ) 3 2 ) . This work extends reinforcement learning to handle non-cumulative objectives.
problem Optimizing functions of rewards rather than their sum in decision processes.
method Mapping NCMDPs to standard MDPs for reinforcement learning.
result Reinforcement learning techniques can be applied to NCMDPs.
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.