This paper explores the integration of strategic optimization methods in search advertising, focusing on ad ranking and bidding mechanisms within E-commerce platforms. By employing a combination of reinforcement learning and evolutionary strategies, we propose a dynamic model that adjusts to varying user interactions and optimizes the balance between advertiser cost, user relevance, and platform revenue. Our results suggest significant improvements in ad placement accuracy and cost efficiency, demonstrating the model's applicability in real-world scenarios.

While originally developed for continuous control problems, Proximal Policy Optimization (PPO) has emerged as the work-horse of a variety of reinforcement learning (RL) applications, including the fine-tuning of generative models. Unfortunately, PPO requires multiple heuristics to enable stable convergence (e.g. value networks, clipping), and is notorious for its sensitivity to the precise implementation of these components. In response, we take a step back and ask what a minimalist RL algorithm for the era of generative models would look like. We propose REBEL, an algorithm that cleanly reduces the problem of policy optimization to regressing the relative reward between two completions to a prompt in terms of the policy, enabling strikingly lightweight implementation. In theory, we prove that fundamental RL algorithms like Natural Policy Gradient can be seen as variants of REBEL, which allows us to match the strongest known theoretical guarantees in terms of convergence and sample complexity in the RL literature. REBEL can also cleanly incorporate offline data and be extended to handle the intransitive preferences we frequently see in practice. Empirically, we find that REBEL provides a unified approach to language modeling and image generation with stronger or similar performance as PPO and DPO, all while being simpler to implement and more computationally efficient than PPO. When fine-tuning Llama-3-8B-Instruct, REBEL achieves strong performance in AlpacaEval 2.0, MT-Bench, and Open LLM Leaderboard.

We explore a Federated Reinforcement Learning (FRL) problem where $N$ agents collaboratively learn a common policy without sharing their trajectory data. To date, existing FRL work has primarily focused on agents operating in the same or ``similar" environments. In contrast, our problem setup allows for arbitrarily large levels of environment heterogeneity. To obtain the optimal policy which maximizes the average performance across all potentially completely different environments, we propose two algorithms: FedSVRPG-M and FedHAPG-M. In contrast to existing results, we demonstrate that both FedSVRPG-M and FedHAPG-M, both of which leverage momentum mechanisms, can exactly converge to a stationary point of the average performance function, regardless of the magnitude of environment heterogeneity. Furthermore, by incorporating the benefits of variance-reduction techniques or Hessian approximation, both algorithms achieve state-of-the-art convergence results, characterized by a sample complexity of $\mathcal{O}\left(\epsilon^{-\frac{3}{2}}/N\right)$. Notably, our algorithms enjoy linear convergence speedups with respect to the number of agents, highlighting the benefit of collaboration among agents in finding a common policy.

Surgical robot task automation has been increasingly Furthermore, the introduction of task-specific rewards and studied for its potential to improve surgical efficiency and the loss of cross-task pretraining create varying internal augment robot intelligence. Recent advancements have witnessed dynamics across tasks, resulting in technical challenges in research on learning-based methods [1]-[5] to promote developing a unified framework for reasoning and decisionmaking automation of surgical robots. Still, current performances within the goal-reaching paradigm in surgical tasks. of the latest methods are impeded in long-horizon To leverage the advanced GPT-based decision-making goal-conditioned tasks, where a sequence of actions and substeps frameworks for improving surgical robot task automation, are required until reaching an ultimate goal. Previous we propose the goal-conditioned decision transformer that algorithms with reinforcement learning [6] and Markov decision embedds goal and time-to-goal as future indicators. Besides, process only predict actions from the current state while we formulate multiple training objectives: action prediction, overlooking information from historical sequential states and dynamics prediction, time-to-goal prediction, and sequence actions. This lacks temporal reasoning capability over actions reconstruction in our cross-task pretraining process, which and affects learning of the inherent sequential dynamics fosters a comprehensive representation of the temporal dynamics which is useful to the final success of a complex task. Despite inherent in goal-conditioned tasks and encourages some works [7], [8] combining task-specific strategies to the model to incorporate diverse temporal reasoning factors.

Language model (LM) post-training (or alignment) involves maximizing a reward function that is derived from preference annotations. Direct Preference Optimization (DPO) is a popular offline alignment method that trains a policy directly on preference data without the need to train a reward model or apply reinforcement learning. However, typical preference datasets have only a single, or at most a few, annotation per preference pair, which causes DPO to overconfidently assign rewards that trend towards infinite magnitude. This frequently leads to degenerate policies, sometimes causing even the probabilities of the preferred generations to go to zero. In this work, we analyze this phenomenon and propose distillation to get a better proxy for the true preference distribution over generation pairs: we train the LM to produce probabilities that match the distribution induced by a reward model trained on the preference data. Moreover, to account for uncertainty in the reward model we are distilling from, we optimize against a family of reward models that, as a whole, is likely to include at least one reasonable proxy for the preference distribution. Our results show that distilling from such a family of reward models leads to improved robustness to distribution shift in preference annotations, while preserving the simple supervised nature of DPO.

We study non-episodic Reinforcement Learning for Lipschitz MDPs in which state-action space is a metric space, and the transition kernel and rewards are Lipschitz functions. We develop computationally efficient UCB-based algorithm, $\textit{ZoRL-}\epsilon$ that adaptively discretizes the state-action space and show that their regret as compared with $\epsilon$-optimal policy is bounded as $\mathcal{O}(\epsilon^{-(2 d_\mathcal{S} + d^\epsilon_z + 1)}\log{(T)})$, where $d^\epsilon_z$ is the $\epsilon$-zooming dimension. In contrast, if one uses the vanilla $\textit{UCRL-}2$ on a fixed discretization of the MDP, the regret w.r.t. a $\epsilon$-optimal policy scales as $\mathcal{O}(\epsilon^{-(2 d_\mathcal{S} + d + 1)}\log{(T)})$ so that the adaptivity gains are huge when $d^\epsilon_z \ll d$. Note that the absolute regret of any 'uniformly good' algorithm for a large family of continuous MDPs asymptotically scales as at least $\Omega(\log{(T)})$. Though adaptive discretization has been shown to yield $\mathcal{\tilde{O}}(H^{2.5}K^\frac{d_z + 1}{d_z + 2})$ regret in episodic RL, an attempt to extend this to the non-episodic case by employing constant duration episodes whose duration increases with $T$, is futile since $d_z \to d$ as $T \to \infty$. The current work shows how to obtain adaptivity gains for non-episodic RL. The theoretical results are supported by simulations on two systems where the performance of $\textit{ZoRL-}\epsilon$ is compared with that of '$\textit{UCRL-C}$,' the fixed discretization-based extension of $\textit{UCRL-}2$ for systems with continuous state-action spaces.

Reinforcement learning (RL) training is inherently unstable due to factors such as moving targets and high gradient variance. Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning from AI Feedback (RLAIF) can introduce additional difficulty. Differing preferences can complicate the alignment process, and prediction errors in a trained reward model can become more severe as the LLM generates unseen outputs. To enhance training robustness, RL has adopted techniques from supervised learning, such as ensembles and layer normalization. In this work, we improve the stability of RL training by adapting the reverse cross entropy (RCE) from supervised learning for noisy data to define a symmetric RL loss. We demonstrate performance improvements across various tasks and scales. We conduct experiments in discrete action tasks (Atari games) and continuous action space tasks (MuJoCo benchmark and Box2D) using Symmetric A2C (SA2C) and Symmetric PPO (SPPO), with and without added noise with especially notable performance in SPPO across different hyperparameters. Furthermore, we validate the benefits of the symmetric RL loss when using SPPO for large language models through improved performance in RLHF tasks, such as IMDB positive sentiment sentiment and TL;DR summarization tasks.

It is becoming increasingly common to deploy algorithms for reinforcement learning and control in systems where the underlying ("latent") dynamics are nonlinear, continuous, and low-dimensional, yet the agent perceives the environment through high-dimensional ("rich") observations such as images from a camera (Wahlström et al., 2015; Levine et al., 2016; Kumar et al., 2021; Nair et al., 2023; Baker et al., 2022; Brohan et al., 2022). These domains demand that agents (i) efficiently explore in the face of complex nonlinearities, and (ii) learn continuous representations that respect the structure of the latent dynamics, ideally in tandem with exploration. In spite of extensive empirical investigation into modeling and algorithm design (Laskin et al., 2020; Yarats et al., 2021a; Hafner et al., 2023), sample-efficiency and reliability remain major challenges (Dean et al., 2020), and our understanding of fundamental algorithmic principles for representation learning and exploration is still in its infancy. Toward understanding algorithmic principles and fundamental limits for reinforcement learning and control with high-dimensional observations, a recent line of theoretical research adopts the framework of richobservation reinforcement learning (c.f., Du et al., 2019; Misra et al., 2020; Mhammedi et al., 2020; Zhang et al., 2022; Mhammedi et al., 2023b). Rich-observation RL provides a mathematical framework for the design and analysis of algorithms that perform exploration in the presence of high-dimensional observations, with an emphasis on generalization and sample-efficiency. However, existing work in this domain is largely restricted to systems with discrete ("tabular") latent dynamics, which is unsuitable for most real-world control applications.

Acquiring a multi-task imitation policy in 3D manipulation poses challenges in terms of scene understanding and action prediction. Current methods employ both 3D representation and multi-view 2D representation to predict the poses of the robot's end-effector. However, they still require a considerable amount of high-quality robot trajectories, and suffer from limited generalization in unseen tasks and inefficient execution in long-horizon reasoning. In this paper, we propose SAM-E, a novel architecture for robot manipulation by leveraging a vision-foundation model for generalizable scene understanding and sequence imitation for long-term action reasoning. Specifically, we adopt Segment Anything (SAM) pre-trained on a huge number of images and promptable masks as the foundation model for extracting task-relevant features, and employ parameter-efficient fine-tuning on robot data for a better understanding of embodied scenarios. To address long-horizon reasoning, we develop a novel multi-channel heatmap that enables the prediction of the action sequence in a single pass, notably enhancing execution efficiency. Experimental results from various instruction-following tasks demonstrate that SAM-E achieves superior performance with higher execution efficiency compared to the baselines, and also significantly improves generalization in few-shot adaptation to new tasks.

Smart meters play a crucial role in enhancing energy management and efficiency, but they raise significant privacy concerns by potentially revealing detailed user behaviors through energy consumption patterns. Recent scholarly efforts have focused on developing battery-aided load-shaping techniques to protect user privacy while balancing costs. This paper proposes a novel deep reinforcement learning-based load-shaping algorithm (PLS-DQN) designed to protect user privacy by proactively creating artificial load signatures that mislead potential attackers. We evaluate our proposed algorithm against a non-intrusive load monitoring (NILM) adversary. The results demonstrate that our approach not only effectively conceals real energy usage patterns but also outperforms state-of-the-art methods in enhancing user privacy while maintaining cost efficiency.