PyTorchKR Papers

최신 논문

총 13,745편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2502.12272

Learning to Reason at the Frontier of Learnability

Reinforcement learning is now widely adopted as the final stage of large language model training, especially for reasoning-style tasks such as maths problems. Typically, models attempt each question many times during a single training step and attempt to learn from their successes and failures. However, we demonstrate that throughout training with two popular algorithms (PPO and VinePPO) on two widely used datasets, many questions are either solved by all attempts - meaning they are already learned - or by none - providing no meaningful training signal. To address this, we adapt a method from the reinforcement learning literature - sampling for learnability - and apply it to the reinforcement learning stage of LLM training. Our curriculum prioritises questions with high variance of success, i.e. those where the agent sometimes succeeds, but not always. Our findings demonstrate that this curriculum consistently boosts training performance across multiple algorithms and datasets, paving the way for more efficient and effective reinforcement learning in LLMs.

arXiv:2502.14200

Causal Mean Field Multi-Agent Reinforcement Learning

Scalability remains a challenge in multi-agent reinforcement learning and is currently under active research. A framework named mean-field reinforcement learning (MFRL) could alleviate the scalability problem by employing the Mean Field Theory to turn a many-agent problem into a two-agent problem. However, this framework lacks the ability to identify essential interactions under nonstationary environments. Causality contains relatively invariant mechanisms behind interactions, though environments are nonstationary. Therefore, we propose an algorithm called causal mean-field Q-learning (CMFQ) to address the scalability problem. CMFQ is ever more robust toward the change of the number of agents though inheriting the compressed representation of MFRL's action-state space. Firstly, we model the causality behind the decision-making process of MFRL into a structural causal model (SCM). Then the essential degree of each interaction is quantified via intervening on the SCM. Furthermore, we design the causality-aware compact representation for behavioral information of agents as the weighted sum of all behavioral information according to their causal effects. We test CMFQ in a mixed cooperative-competitive game and a cooperative game. The result shows that our method has excellent scalability performance in both training in environments containing a large number of agents and testing in environments containing much more agents.

arXiv:2502.14307

μRL: Discovering Transient Execution Vulnerabilities Using Reinforcement Learning

We propose using reinforcement learning to address the challenges of discovering microarchitectural vulnerabilities, such as Spectre and Meltdown, which exploit subtle interactions in modern processors. Traditional methods like random fuzzing fail to efficiently explore the vast instruction space and often miss vulnerabilities that manifest under specific conditions. To overcome this, we introduce an intelligent, feedback-driven approach using RL. Our RL agents interact with the processor, learning from real-time feedback to prioritize instruction sequences more likely to reveal vulnerabilities, significantly improving the efficiency of the discovery process. We also demonstrate that RL systems adapt effectively to various microarchitectures, providing a scalable solution across processor generations. By automating the exploration process, we reduce the need for human intervention, enabling continuous learning that uncovers hidden vulnerabilities. Additionally, our approach detects subtle signals, such as timing anomalies or unusual cache behavior, that may indicate microarchitectural weaknesses. This proposal advances hardware security testing by introducing a more efficient, adaptive, and systematic framework for protecting modern processors. When unleashed on Intel Skylake-X and Raptor Lake microarchitectures, our RL agent was indeed able to generate instruction sequences that cause significant observable byte leakages through transient execution without generating any $\mu$code assists, faults or interrupts. The newly identified leaky sequences stem from a variety of Intel instructions, e.g. including SERIALIZE, VERR/VERW, CLMUL, MMX-x87 transitions, LSL+RDSCP and LAR. These initial results give credence to the proposed approach.

arXiv:2502.13137

Theorem Prover as a Judge for Synthetic Data Generation

The demand for synthetic data in mathematical reasoning has increased due to its potential to enhance the mathematical capabilities of large language models (LLMs). However, ensuring the validity of intermediate reasoning steps remains a significant challenge, affecting data quality. While formal verification via theorem provers effectively validates LLM reasoning, the autoformalisation of mathematical proofs remains error-prone. In response, we introduce iterative autoformalisation, an approach that iteratively refines theorem prover formalisation to mitigate errors, thereby increasing the execution rate on the Lean prover from 60% to 87%. Building upon that, we introduce Theorem Prover as a Judge (TP-as-a-Judge), a method that employs theorem prover formalisation to rigorously assess LLM intermediate reasoning, effectively integrating autoformalisation with synthetic data generation. Finally, we present Reinforcement Learning from Theorem Prover Feedback (RLTPF), a framework that replaces human annotation with theorem prover feedback in Reinforcement Learning from Human Feedback (RLHF). Across multiple LLMs, applying TP-as-a-Judge and RLTPF improves benchmarks with only 3,508 samples, achieving 5.56% accuracy gain on Mistral-7B for MultiArith, 6.00% on Llama-2-7B for SVAMP, and 3.55% on Llama-3.1-8B for AQUA.

arXiv:2502.12876

Continuous Learning Conversational AI: A Personalized Agent Framework via A2C Reinforcement Learning

Creating personalized and adaptable conversational AI remains a key challenge. This paper introduces a Continuous Learning Conversational AI (CLCA) approach, implemented using A2C reinforcement learning, to move beyond static Large Language Models (LLMs). We use simulated sales dialogues, generated by LLMs, to train an A2C agent. This agent learns to optimize conversation strategies for personalization, focusing on engagement and delivering value. Our system architecture integrates reinforcement learning with LLMs for both data creation and response selection. This method offers a practical way to build personalized AI companions that evolve through continuous learning, advancing beyond traditional static LLM techniques.

arXiv:2502.13822

Uncertainty quantification for Markov chains with application to temporal difference learning

Markov chains are fundamental to statistical machine learning, underpinning key methodologies such as Markov Chain Monte Carlo (MCMC) sampling and temporal difference (TD) learning in reinforcement learning (RL). Given their widespread use, it is crucial to establish rigorous probabilistic guarantees on their convergence, uncertainty, and stability. In this work, we develop novel, high-dimensional concentration inequalities and Berry-Esseen bounds for vector- and matrix-valued functions of Markov chains, addressing key limitations in existing theoretical tools for handling dependent data. We leverage these results to analyze the TD learning algorithm, a widely used method for policy evaluation in RL. Our analysis yields a sharp high-probability consistency guarantee that matches the asymptotic variance up to logarithmic factors. Furthermore, we establish a $O(T^{-\frac{1}{4}}\log T)$ distributional convergence rate for the Gaussian approximation of the TD estimator, measured in convex distance. These findings provide new insights into statistical inference for RL algorithms, bridging the gaps between classical stochastic approximation theory and modern reinforcement learning applications.

arXiv:2502.12756공식 코드 · Papers with Code

Navigating Demand Uncertainty in Container Shipping: Deep Reinforcement Learning for Enabling Adaptive and Feasible Master Stowage Planning

Reinforcement learning (RL) has shown promise in solving various combinatorial optimization problems. However, conventional RL faces challenges when dealing with real-world constraints, especially when action space feasibility is explicit and dependent on the corresponding state or trajectory. In this work, we focus on using RL in container shipping, often considered the cornerstone of global trade, by dealing with the critical challenge of master stowage planning. The main objective is to maximize cargo revenue and minimize operational costs while navigating demand uncertainty and various complex operational constraints, namely vessel capacity and stability, which must be dynamically updated along the vessel's voyage. To address this problem, we implement a deep reinforcement learning framework with feasibility projection to solve the master stowage planning problem (MPP) under demand uncertainty. The experimental results show that our architecture efficiently finds adaptive, feasible solutions for this multi-stage stochastic optimization problem, outperforming traditional mixed-integer programming and RL with feasibility regularization. Our AI-driven decision-support policy enables adaptive and feasible planning under uncertainty, optimizing operational efficiency and capacity utilization while contributing to sustainable and resilient global supply chains.

arXiv:2502.13584

Multi-Target Radar Search and Track Using Sequence-Capable Deep Reinforcement Learning

The research addresses sensor task management for radar systems, focusing on efficiently searching and tracking multiple targets using reinforcement learning. The approach develops a 3D simulation environment with an active electronically scanned array radar, using a multi-target tracking algorithm to improve observation data quality. Three neural network architectures were compared including an approach using fated recurrent units with multi-headed self-attention. Two pre-training techniques were applied: behavior cloning to approximate a random search strategy and an auto-encoder to pre-train the feature extractor. Experimental results revealed that search performance was relatively consistent across most methods. The real challenge emerged in simultaneously searching and tracking targets. The multi-headed self-attention architecture demonstrated the most promising results, highlighting the potential of sequence-capable architectures in handling dynamic tracking scenarios. The key contribution lies in demonstrating how reinforcement learning can optimize sensor management, potentially improving radar systems' ability to identify and track multiple targets in complex environments.

arXiv:2502.12624

Implicit Repair with Reinforcement Learning in Emergent Communication

Conversational repair is a mechanism used to detect and resolve miscommunication and misinformation problems when two or more agents interact. One particular and underexplored form of repair in emergent communication is the implicit repair mechanism, where the interlocutor purposely conveys the desired information in such a way as to prevent misinformation from any other interlocutor. This work explores how redundancy can modify the emergent communication protocol to continue conveying the necessary information to complete the underlying task, even with additional external environmental pressures such as noise. We focus on extending the signaling game, called the Lewis Game, by adding noise in the communication channel and inputs received by the agents. Our analysis shows that agents add redundancy to the transmitted messages as an outcome to prevent the negative impact of noise on the task success. Additionally, we observe that the emerging communication protocol's generalization capabilities remain equivalent to architectures employed in simpler games that are entirely deterministic. Additionally, our method is the only one suitable for producing robust communication protocols that can handle cases with and without noise while maintaining increased generalization performance levels.

arXiv:2502.11968

Theoretical Barriers in Bellman-Based Reinforcement Learning

Reinforcement Learning algorithms designed for high-dimensional spaces often enforce the Bellman equation on a sampled subset of states, relying on generalization to propagate knowledge across the state space. In this paper, we identify and formalize a fundamental limitation of this common approach. Specifically, we construct counterexample problems with a simple structure that this approach fails to exploit. Our findings reveal that such algorithms can neglect critical information about the problems, leading to inefficiencies. Furthermore, we extend this negative result to another approach from the literature: Hindsight Experience Replay learning state-to-state reachability.

arXiv:2502.11937

FitLight: Federated Imitation Learning for Plug-and-Play Autonomous Traffic Signal Control

Although Reinforcement Learning (RL)-based Traffic Signal Control (TSC) methods have been extensively studied, their practical applications still raise some serious issues such as high learning cost and poor generalizability. This is because the ``trial-and-error'' training style makes RL agents extremely dependent on the specific traffic environment, which also requires a long convergence time. To address these issues, we propose a novel Federated Imitation Learning (FIL)-based framework for multi-intersection TSC, named FitLight, which allows RL agents to plug-and-play for any traffic environment without additional pre-training cost. Unlike existing imitation learning approaches that rely on pre-training RL agents with demonstrations, FitLight allows real-time imitation learning and seamless transition to reinforcement learning. Due to our proposed knowledge-sharing mechanism and novel hybrid pressure-based agent design, RL agents can quickly find a best control policy with only a few episodes. Moreover, for resource-constrained TSC scenarios, FitLight supports model pruning and heterogeneous model aggregation, such that RL agents can work on a micro-controller with merely 16{\it KB} RAM and 32{\it KB} ROM. Extensive experiments demonstrate that, compared to state-of-the-art methods, FitLight not only provides a superior starting point but also converges to a better final solution on both real-world and synthetic datasets, even under extreme resource limitations.

arXiv:2502.11612공식 코드 · Papers with Code

Maximum Entropy Reinforcement Learning with Diffusion Policy

The Soft Actor-Critic (SAC) algorithm with a Gaussian policy has become a mainstream implementation for realizing the Maximum Entropy Reinforcement Learning (MaxEnt RL) objective, which incorporates entropy maximization to encourage exploration and enhance policy robustness. While the Gaussian policy performs well on simpler tasks, its exploration capacity and potential performance in complex multi-goal RL environments are limited by its inherent unimodality. In this paper, we employ the diffusion model, a powerful generative model capable of capturing complex multimodal distributions, as the policy representation to fulfill the MaxEnt RL objective, developing a method named MaxEnt RL with Diffusion Policy (MaxEntDP). Our method enables efficient exploration and brings the policy closer to the optimal MaxEnt policy. Experimental results on Mujoco benchmarks show that MaxEntDP outperforms the Gaussian policy and other generative models within the MaxEnt RL framework, and performs comparably to other state-of-the-art diffusion-based online RL algorithms. Our code is available at https://github.com/diffusionyes/MaxEntDP.

arXiv:2502.10792

Tackling the Zero-Shot Reinforcement Learning Loss Directly

Zero-shot reinforcement learning (RL) methods aim at instantly producing a behavior for an RL task in a given environment, from a description of the reward function. These methods are usually tested by evaluating their average performance on a series of downstream tasks. Yet they cannot be trained directly for that objective, unless the distribution of downstream tasks is known. Existing approaches either use other learning criteria [BBQ+ 18, TRO23, TO21, HDB+ 19], or explicitly set a prior on downstream tasks, such as reward functions given by a random neural network [FPAL24]. Here we prove that the zero-shot RL loss can be optimized directly, for a range of non-informative priors such as white noise rewards, temporally smooth rewards, ``scattered'' sparse rewards, or a combination of those. Thus, it is possible to learn the optimal zero-shot features algorithmically, for a wide mixture of priors. Surprisingly, the white noise prior leads to an objective almost identical to the one in VISR [HDB+19], via a different approach. This shows that some seemingly arbitrary choices in VISR, such as Von Mises--Fisher distributions, do maximize downstream performance. This also suggests more efficient ways to tackle the VISR objective. Finally, we discuss some consequences and limitations of the zero-shot RL objective, such as its tendency to produce narrow optimal features if only using Gaussian dense reward priors.

arXiv:2502.10732

Rule-Bottleneck Reinforcement Learning: Joint Explanation and Decision Optimization for Resource Allocation with Language Agents

Deep Reinforcement Learning (RL) is remarkably effective in addressing sequential resource allocation problems in domains such as healthcare, public policy, and resource management. However, deep RL policies often lack transparency and adaptability, challenging their deployment alongside human decision-makers. In contrast, Language Agents, powered by large language models (LLMs), provide human-understandable reasoning but may struggle with effective decision making. To bridge this gap, we propose Rule-Bottleneck Reinforcement Learning (RBRL), a novel framework that jointly optimizes decision and explanations. At each step, RBRL generates candidate rules with an LLM, selects among them using an attention-based RL policy, and determines the environment action with an explanation via chain-of-thought reasoning. The RL rule selection is optimized using the environment rewards and an explainability metric judged by the LLM. Evaluations in real-world scenarios highlight RBRL's competitive performance with deep RL and efficiency gains over LLM fine-tuning. A survey further confirms the enhanced quality of its explanations.

arXiv:2502.10431공식 코드 · Papers with Code

Leveraging Constraint Violation Signals For Action-Constrained Reinforcement Learning

In many RL applications, ensuring an agent's actions adhere to constraints is crucial for safety. Most previous methods in Action-Constrained Reinforcement Learning (ACRL) employ a projection layer after the policy network to correct the action. However projection-based methods suffer from issues like the zero gradient problem and higher runtime due to the usage of optimization solvers. Recently methods were proposed to train generative models to learn a differentiable mapping between latent variables and feasible actions to address this issue. However, generative models require training using samples from the constrained action space, which itself is challenging. To address such limitations, first, we define a target distribution for feasible actions based on constraint violation signals, and train normalizing flows by minimizing the KL divergence between an approximated distribution over feasible actions and the target. This eliminates the need to generate feasible action samples, greatly simplifying the flow model learning. Second, we integrate the learned flow model with existing deep RL methods, which restrict it to exploring only the feasible action space. Third, we extend our approach beyond ACRL to handle state-wise constraints by learning the constraint violation signal from the environment. Empirically, our approach has significantly fewer constraint violations while achieving similar or better quality in several control tasks than previous best methods.

arXiv:2502.10867

A Tutorial on LLM Reasoning: Relevant Methods behind ChatGPT o1

OpenAI o1 has shown that applying reinforcement learning to integrate reasoning steps directly during inference can significantly improve a model's reasoning capabilities. This result is exciting as the field transitions from the conventional autoregressive method of generating answers to a more deliberate approach that models the slow-thinking process through step-by-step reasoning training. Reinforcement learning plays a key role in both the model's training and decoding processes. In this article, we present a comprehensive formulation of reasoning problems and investigate the use of both model-based and model-free approaches to better support this slow-thinking framework.

arXiv:2502.11134

Solving Online Resource-Constrained Scheduling for Follow-Up Observation in Astronomy: a Reinforcement Learning Approach

In the astronomical observation field, determining the allocation of observation resources of the telescope array and planning follow-up observations for targets of opportunity (ToOs) are indispensable components of astronomical scientific discovery. This problem is computationally challenging, given the online observation setting and the abundance of time-varying factors that can affect whether an observation can be conducted. This paper presents ROARS, a reinforcement learning approach for online astronomical resource-constrained scheduling. To capture the structure of the astronomical observation scheduling, we depict every schedule using a directed acyclic graph (DAG), illustrating the dependency of timing between different observation tasks within the schedule. Deep reinforcement learning is used to learn a policy that can improve the feasible solution by iteratively local rewriting until convergence. It can solve the challenge of obtaining a complete solution directly from scratch in astronomical observation scenarios, due to the high computational complexity resulting from numerous spatial and temporal constraints. A simulation environment is developed based on real-world scenarios for experiments, to evaluate the effectiveness of our proposed scheduling approach. The experimental results show that ROARS surpasses 5 popular heuristics, adapts to various observation scenarios and learns effective strategies with hindsight.

arXiv:2502.10456

Deep Reinforcement Learning-Based User Scheduling for Collaborative Perception

Stand-alone perception systems in autonomous driving suffer from limited sensing ranges and occlusions at extended distances, potentially resulting in catastrophic outcomes. To address this issue, collaborative perception is envisioned to improve perceptual accuracy by using vehicle-to-everything (V2X) communication to enable collaboration among connected and autonomous vehicles and roadside units. However, due to limited communication resources, it is impractical for all units to transmit sensing data such as point clouds or high-definition video. As a result, it is essential to optimize the scheduling of communication links to ensure efficient spectrum utilization for the exchange of perceptual data. In this work, we propose a deep reinforcement learning-based V2X user scheduling algorithm for collaborative perception. Given the challenges in acquiring perceptual labels, we reformulate the conventional label-dependent objective into a label-free goal, based on characteristics of 3D object detection. Incorporating both channel state information (CSI) and semantic information, we develop a double deep Q-Network (DDQN)-based user scheduling framework for collaborative perception, named SchedCP. Simulation results verify the effectiveness and robustness of SchedCP compared with traditional V2X scheduling methods. Finally, we present a case study to illustrate how our proposed algorithm adaptively modifies the scheduling decisions by taking both instantaneous CSI and perceptual semantics into account.

arXiv:2502.10429

Real Time Control of Tandem-Wing Experimental Platform Using Concerto Reinforcement Learning

This paper introduces the CRL2RT algorithm, an advanced reinforcement learning method aimed at improving the real-time control performance of the Direct-Drive Tandem-Wing Experimental Platform (DDTWEP). Inspired by dragonfly flight, DDTWEP's tandem wing structure causes nonlinear and unsteady aerodynamic interactions, leading to complex load behaviors during pitch, roll, and yaw maneuvers. These complexities challenge stable motion control at high frequencies (2000 Hz). To overcome these issues, we developed the CRL2RT algorithm, which combines classical control elements with reinforcement learning-based controllers using a time-interleaved architecture and a rule-based policy composer. This integration ensures finite-time convergence and single-life adaptability. Experimental results under various conditions, including different flapping frequencies and yaw disturbances, show that CRL2RT achieves a control frequency surpassing 2500 Hz on standard CPUs. Additionally, when integrated with classical controllers like PID, Adaptive PID, and Model Reference Adaptive Control (MRAC), CRL2RT enhances tracking performance by 18.3% to 60.7%. These findings demonstrate CRL2RT's broad applicability and superior performance in complex real-time control scenarios, validating its effectiveness in overcoming existing control strategy limitations and advancing robust, efficient real-time control for biomimetic aerial vehicles.

arXiv:2502.10187

Reinforcement Learning based Constrained Optimal Control: an Interpretable Reward Design

This paper presents an interpretable reward design framework for reinforcement learning based constrained optimal control problems with state and terminal constraints. The problem is formalized within a standard partially observable Markov decision process framework. The reward function is constructed from four weighted components: a terminal constraint reward, a guidance reward, a penalty for state constraint violations, and a cost reduction incentive reward. A theoretically justified reward design is then presented, which establishes bounds on the weights of the components. This approach ensures that constraints are satisfied and objectives are optimized while mitigating numerical instability. Acknowledging the importance of prior knowledge in reward design, we sequentially solve two subproblems, using each solution to inform the reward design for the subsequent problem. Subsequently, we integrate reinforcement learning with curriculum learning, utilizing policies derived from simpler subproblems to assist in tackling more complex challenges, thereby facilitating convergence. The framework is evaluated against original and randomly weighted reward designs in a multi-agent particle environment. Experimental results demonstrate that the proposed approach significantly enhances satisfaction of terminal and state constraints and optimization of control cost.