The paper introduces Loopless Variance-Reduced Policy Gradient (LVR-PG), a framework that unifies first‑order estimators combining recursive updates with defensive importance sampling, such as PAGE-PG and STORM-PG. It proves that LVR-PG achieves a ∫O(ε⁻²) last‑iterate global convergence rate under weak gradient domination for general parametric policies, using only first‑order updates and minimal assumptions. By extending lower bounds to continuous parametric policies, the authors show an Ω(ε⁻²) sample complexity limit, establishing that LVR-PG is minimax‑optimal up to logarithmic factors and horizon terms.
arXiv:2506. 06178v3 Announce Type: replace Abstract: Policy gradient (PG) methods are a class of effective reinforcement learning algorithms, particularly when dealing with continuous control problems.
arXiv:2610.03165v1 Announce Type: new Abstract: Several variance-reduced versions of REINFORCE based on importance sampling achieve an improved $O(\epsilon^{-3})$ sample complexity to find an $\epsil...
The paper investigates best‑policy identification in finite‑horizon, risk‑sensitive reinforcement learning using the entropic risk measure. It identifies a gap between known lower bounds ≥ η(e^{|eta|H}) and upper bounds ≤ O(e^{2|eta|H}) for sample complexity, attributing the excess factor to loose concentration bounds for exponential utilities. By employing a forward‑model algorithm with KL‑based exploration bonuses and a novel stopping rule, the authors achieve a sample complexity that matches the lower bound, closing the previously open exponential gap.
arXiv:2503.22244v3 Announce Type: replace Abstract: Policy gradient methods are among the most successful approaches for solving challenging reinforcement learning problems. Despite their empirical s...
arXiv:2512. 23075v5 Announce Type: replace-cross Abstract: Policy gradient methods for Large Language Models optimize a policy $\pi_\theta$ via a surrogate objective computed from samples of a rollout policy $\pi_{\text{roll}}$.
The paper introduces Fast Regularized Policy Mirror Descent (PMD) that pairs policy updates with a single temporal-difference (TD) critic step. It proves global linear convergence for finite discounted MDPs using exact coordinate-wise Bellman updates and any positive actor stepsize, regardless of critic initialization. For stochastic TD-PMD with strongly convex mirror maps, the authors achieve an expected value gap of ε after “~O(1/((1-γ)^5 σ_b ε))” transitions, without requiring trajectory resets, generative models, or nested policy evaluation loops.
arXiv:2607. 28390v1 Announce Type: new Abstract: Constrained Markov Decision Processes (CMDPs) provide a natural framework for reinforcement learning in safety-critical applications, where agents maximize long-term reward while satisfying long-term constraints.
arXiv:2602. 09300v2 Announce Type: replace Abstract: We consider the policy evaluation and control in a finite horizon reinforcement learning (RL) setting under an expectile-based objective.
The paper presents a policy gradient theorem tailored to Cumulative Prospect Theory (CPT) objectives in finite-horizon reinforcement learning, extending the classic policy gradient framework to include distortion-based risk measures. Leveraging this theorem, the authors develop a first‑order policy gradient algorithm that uses a Monte Carlo estimator based on order statistics, providing statistical guarantees and proving asymptotic convergence to first‑order stationary points of the generally nonconvex CPT objective. The work also offers a non‑asymptotic sample complexity bound for reaching an approximate stationary policy and demonstrates the qualitative effects of CPT through simulations, comparing the new first‑order method to existing zeroth‑order approaches.
The paper investigates stochastic contextual bandits in an agnostic setting, aiming to compete with the best policy in a given class without assuming realizability or specific loss/reward models. It introduces an algorithm that updates the policy each round by minimizing a pessimistic objective— a clipped inverse‑propensity estimate of the policy value plus a variance penalty— and proves the first fast regret rates relative to the best‑in‑class policy. By exploiting entropy assumptions on the policy class and a H"olderian error‑bound condition, the authors achieve fast best‑in‑class regret rates, including polylogarithmic rates in the parametric case, using a sequential self‑normalized maximal inequality for bounded martingale empirical processes to derive uniform variance‑adaptive confidence bounds and ensure pessimism under adaptive data collection.
The paper introduces Batched SGD, a variant that groups online samples into epochs and performs a single update per epoch using a low‑variance gradient estimate. This batching approach allows a straightforward high‑probability analysis without restrictive assumptions or auxiliary sequences, yielding near‑optimal rates for both strongly convex and non‑convex objectives under standard smoothness and sub‑Gaussian noise conditions. The authors also extend the method to federated learning, providing the first high‑probability guarantees with logarithmic communication complexity, linear speedup in the number of agents, and robustness to data heterogeneity.
The paper introduces Generalized Proximal Policy Optimization (GPPO), a unified framework that extends PPO by incorporating $f$-divergence trust-region theory and novel estimator topologies for large language model reinforcement learning. GPPO demonstrates that different divergences produce distinct policy update dynamics and that estimator design significantly impacts performance. Experiments on eight math and coding benchmarks show GPPO surpasses standard GRPO by up to 50% in math and 14% in coding, achieving state‑of‑the‑art results.