This paper deals with an important problem in theoretical reinforcement learning (RL), that is, finite-time analysis of on-policy RL algorithms such as SARSA. If the analysis techniques, as well as proofs, were correct and concrete, this work may have a broad impact on analyzing related stochastic approximation/RL algorithms. Although important and interesting, the present submission contains several major concerns, that have limited the contributions and even brought into question the practical usefulness of the reported theoretical results. These concerns are listed as follows. To facilitate analysis, a number of the assumptions adopted in this work are strong and impractical.

The paper focuses on the important problem of designing optimal algorithms for exploration-exploitation (whose upper-bound matches the lower bound). The paper is not well organized and written. It is difficult to abstract from the mathematical formulation and grasps the key ideas behind the improvement of the regret bound. As far as I understood, the first important component in improving the bound is to use variance dependent confidence intervals (ie Bernstein). Together with the knowledge of H, this allows designing a tighter optimism (Eq.

This paper has lead to a long and thoughtful discussion between the reviewers. The main points that were raised are the following: The results are novel and close a long-standing gap between upper and lower bounds in a very important problem. While the reviewers have agreed that the results are significant and they definitely bring the field forward, an expert reviewer argued that the step forward is perhaps not significantly big enough to warrant publication in the present form. However, after much discussion, the other reviewers made a strong case for acceptance and all reviewers agreed that the community would clearly benefit from this paper being published. That said, I strongly encourage the authors to work hard on improving the presentation for the final version.

Additional Feedback: I like authors tried their experiments in various perspectives, but experience sharing is occasionally seen from the existing literature. For example, although it wasn't mentioned in the paper, [1] used experience sharing among agents for their implementation, and I believe there may be other works with the topic of "MARL for homogeneous agents". The main reason I score "below acceptance" is that quite weak baselines seem to be used: - In Table 1, QMIX and MADDPG highly underperforms SEAC and other baselines (IAC, SNAC). However, since methods with CTDE are mostly more stable than independent learning methods, I think this part should be explained in more detail. Although other reviewers have argued the strength of this work from the importance weighting and simplicity of methods, I still think there should have been stronger baselines.

This paper introduces a simple idea for MARL, using importance weights to correct for off-policy. Generally, the reviewers agree that the paper is clear and well written. Although the main idea is very natural and intuitive, as pointed out by reviewer 4, it is not intuitive that is would actually work. Therefore, one of the strengths of this paper is to show that intuition fails us in this case. The reviewers point out some weaknesses in the empirical sections, in particular comparisons with other methods, and we hope that the authors will be able to address some of these in the final version of the paper.

Reinforcement Learning (RL) enables an intelligent agent to optimise its performance in a task by continuously taking action from an observed state and receiving a feedback from the environment in form of rewards. RL typically uses tables or linear approximators to map state-action tuples that maximises the reward. Combining RL with deep neural networks (DRL) significantly increases its scalability and enables it to address more complex problems than before. However, DRL also inherits downsides from both RL and deep learning. Despite DRL improves generalisation across similar state-action pairs when compared to simpler RL policy representations like tabular methods, it still requires the agent to adequately explore the state-action space. Additionally, deep methods require more training data, with the volume of data escalating with the complexity and size of the neural network. As a result, deep RL requires a long time to collect enough agent-environment samples and to successfully learn the underlying policy. Furthermore, often even a slight alteration to the task invalidates any previous acquired knowledge. To address these shortcomings, Transfer Learning (TL) has been introduced, which enables the use of external knowledge from other tasks or agents to enhance a learning process. The goal of TL is to reduce the learning complexity for an agent dealing with an unfamiliar task by simplifying the exploration process. This is achieved by lowering the amount of new information required by its learning model, resulting in a reduced overall convergence time...

In online advertising, advertisers commonly utilize auto-bidding services to bid for impression opportunities. A typical objective of the auto-bidder is to optimize the advertiser's cumulative value of winning impressions within specified budget constraints. However, such a problem is challenging due to the complex bidding environment faced by diverse advertisers. To address this challenge, we introduce ABPlanner, a few-shot adaptable budget planner designed to improve budget-constrained auto-bidding. ABPlanner is based on a hierarchical bidding framework that decomposes the bidding process into shorter, manageable stages. Within this framework, ABPlanner allocates the budget across all stages, allowing a low-level auto-bidder to bids based on the budget allocation plan. The adaptability of ABPlanner is achieved through a sequential decision-making approach, inspired by in-context reinforcement learning. For each advertiser, ABPlanner adjusts the budget allocation plan episode by episode, using data from previous episodes as prompt for current decisions. This enables ABPlanner to quickly adapt to different advertisers with few-shot data, providing a sample-efficient solution. Extensive simulation experiments and real-world A/B testing validate the effectiveness of ABPlanner, demonstrating its capability to enhance the cumulative value achieved by auto-bidders.

Exploration in sparse reward environments remains a significant challenge in reinforcement learning, particularly in Contextual Markov Decision Processes (CMDPs), where environments differ across episodes. Existing episodic intrinsic motivation methods for CMDPs primarily rely on count-based approaches, which are ineffective in large state spaces, or on similarity-based methods that lack appropriate metrics for state comparison. To address these shortcomings, we propose Episodic Novelty Through Temporal Distance (ETD), a novel approach that introduces temporal distance as a robust metric for state similarity and intrinsic reward computation. By employing contrastive learning, ETD accurately estimates temporal distances and derives intrinsic rewards based on the novelty of states within the current episode. Extensive experiments on various benchmark tasks demonstrate that ETD significantly outperforms state-of-the-art methods, highlighting its effectiveness in enhancing exploration in sparse reward CMDPs.

Decentralized Multi-Agent Reinforcement Learning (Dec-MARL) has emerged as a pivotal approach for addressing complex tasks in dynamic environments. Existing Multi-Agent Reinforcement Learning (MARL) methodologies typically assume a shared objective among agents and rely on centralized control. However, many real-world scenarios feature agents with individual goals and limited observability of other agents, complicating coordination and hindering adaptability. Existing Dec-MARL strategies prioritize either communication or coordination, lacking an integrated approach that leverages both. This paper presents a novel Dec-MARL framework that integrates peer-to-peer communication and coordination, incorporating goal-awareness and time-awareness into the agents' knowledge-sharing processes. Our framework equips agents with the ability to (i) share contextually relevant knowledge to assist other agents, and (ii) reason based on information acquired from multiple agents, while considering their own goals and the temporal context of prior knowledge. We evaluate our approach through several complex multi-agent tasks in environments with dynamically appearing obstacles. Our work demonstrates that incorporating goal-aware and time-aware knowledge sharing significantly enhances overall performance.

We propose a novel multi-agent reinforcement learning (RL) approach for inter-cell interference mitigation, in which agents selectively share their experiences with other agents. Each base station is equipped with an agent, which receives signal-to-interference-plus-noise ratio from its own associated users. This information is used to evaluate and selectively share experiences with neighboring agents. The idea is that even a few pertinent experiences from other agents can lead to effective learning. This approach enables fully decentralized training and execution, minimizes information sharing between agents and significantly reduces communication overhead, which is typically the burden of interference management. The proposed method outperforms state-of-the-art multi-agent RL techniques where training is done in a decentralized manner. Furthermore, with a 75% reduction in experience sharing, the proposed algorithm achieves 98% of the spectral efficiency obtained by algorithms sharing all experiences.