Reinforcement Learning

Recent advances in text-to-speech (TTS) synthesis have achieved highly natural and expressive speech generation. However, these systems are designed for general adults and overlook older adults' speech comprehension needs due to age-related sensory and cognitive decline. Prior work involves older adults by collecting preference feedback to tune model parameters. However, obtaining sufficient preference data is costly and difficult, as older adults quickly become fatigued during collection. In this paper, we propose a novel imitation learning (IL) framework to learn TTS models from expert demonstrations. We further improve Group Relative Policy Optimization (GRPO) with two-stage on-policy reward learning (OPRL) to mitigate reward hacking under limited supervision from expert demonstration. Experimental results show that GRPO w/ OPRL outperforms GRPO and supervised baselines in objective and subjective metrics. Audio samples are available at https://dongru1.github.io/demo/im-efss

Reinforcement learning with verifiable rewards (RLVR) has driven substantial progress in large language model reasoning, but relies on ground-truth supervision that is costly or infeasible, especially in coding tasks. Recent work addresses this by deriving rewards from a model's own signals, such as majority voting or confidence-based scores, achieving notable success on mathematical reasoning benchmarks. However, code generation poses distinct challenges: programs are structurally complex, semantically equivalent solutions may differ syntactically, and verification typically requires execution. Whether these intrinsic reward methods transfer effectively to code remains unexplored. In this work, we present a systematic empirical study of intrinsic reward methods for code generation. We conduct extensive experiments on LiveCodeBench, systematically evaluating representative certainty-based Reinforcement Learning from Internal Feedback (RLIF) approaches under different training scenarios and hyperparameter settings. Our experiments reveal that certainty-based methods yield early gains but inevitably collapse: models progressively shorten outputs and lose reasoning capability, with collapse speed sensitive to sample size and temperature. When used to initialize RLVR training, RLIF pre-training offers no significant improvement over training from scratch. We also provide actionable recommendations for using intrinsic rewards for training code reasoning models. Our study shows both the promise and limitations of intrinsic reward methods for code, informing future work on code models and agents.

The temporal structure of reward composition in reinforcement learning (RL) is typically hand-designed and held fixed throughout training, leaving the progression of motivational priorities largely unexplored. In this work, we propose an evolutionary framework for discovering developmental reward schedules, in which three distinct biologically inspired motivational components -- agency, novelty, and reactivity -- are combined through time-varying weights that dynamically shift over the course of training. Evaluated on two sparse-reward MiniGrid tasks: DoorKey-6x6 and KeyCorridorS3R1, our framework compares the generalizability of four evolutionary algorithms: CMA-ES, xNES, DE, and L-SHADE against an extrinsically motivated baseline (our main comparison point), and three additional hand-designed methods. On DoorKey-6x6, all evolved methods outperform the non-evolved baselines, with L-SHADE achieving the best performance -- an approximate relative mean improvement of 11.4% over the extrinsic only baseline. On KeyCorridorS3R1, CMA-ES achieves the best overall performance, with the remaining evolved methods showing weaker and less reliable generalization capability compared to the extrinsic only baseline. Interestingly, the discovered schedules diverge from our defined developmental ordering, with novelty consistently emerging as the dominant early signal during training, across both tasks. Collectively, our results position evolutionary optimization as a promising approach for developmental reward schedule discovery in deep reinforcement learning, and suggest that what evolution finds to be optimal in computational settings may differ from what it finds to be optimal in biology. The code for this project can be found at: https://github.com/alannadels/Evolutionary_RL.git.

Astrophysical simulations frequently address multi-scale, multi-physics problems through subsystem decomposition, problem-tailored integration schemes, and coupling on fixed manually set timescales. Here we introduce ReLaTS, a reinforcement learning framework that dynamically selects the coupling time step to optimize the trade-off between accuracy and computational cost. We validate ReLaTS on star clusters containing a planetary system, and test the method by varying the number of stars $N_\star$ in the cluster and the number of planets ($N_{\rm planet}$) orbiting one of them. The method finds the optimal coupling time step that balances speed and accuracy without requiring expert knowledge. In addition, the trained network operates independently of the coupled \textit{N}-body algorithms, displaying stable performance across a range of setups. We observe that the method is less reliable for cases with infinitesimal masses, as their contribution to the total energy is negligible compared to that of the massive bodies, and the network is not capable of recognizing potential errors generated while integrating them. For long-time integration of large $N$ systems, the error accumulates. The reinforcement learning algorithm, however, manages to keep the energy error below a pre-set threshold. This approach substantially reduces energy errors relative to fixed-time step baselines without substantial additional computational overhead. Once trained, ReLaTS requires no expert tuning and generalizes across diverse astrophysical domains, enabling adaptive multi-scale simulations.

Safety is a core concern for deploying reinforcement learning (RL) agents in real-world domains such as robotics and autonomous driving. While benchmarks have been central to progress in RL, existing 3D physics-based safety benchmarks remain computationally slow, limiting large-scale experimentation and rapid prototyping. To address this gap, we propose CRAX (Constrained RL Accelerated with JAX). Built on top of the MuJoCo XLA (MJX) physics engine, CRAX leverages vectorized operations and hardware acceleration, yielding up to 200x faster training over comparable CPU-based safety benchmarks. The benchmark features eight tasks spanning three difficulty levels and multiple agent morphologies. Evaluating seven popular safe RL methods, we find that none dominates across tasks, and that learning safe policies from pixels remains largely unsolved.

Virtual training environments are software-intensive systems in which reinforcement learning (RL) agents learn, adapt, and demonstrate meaningful behavior. Virtual training environments offer a safe and cost-efficient alternative to training agents in real-world settings. However, to converge, most realistic RL problems require training in multiple, mostly similar but slightly different environments - i.e., families of environment variants. The typical development process of environment families is a labor-intensive and error-prone manual endeavor that does not scale well. To alleviate these issues, in this paper, we propose a model-driven approach for developing families of RL training environments. To obtain the family of environments, we develop an approach and prototype tool. In our approach, a hybrid genetic algorithm - a combination of population-based global search and heuristic local search - generates environment families. Mutations and constraints are expressed as model transformations and are operationalized into a search process by a state-of-the-art model transformation engine. We demonstrate the soundness of our approach in a wildfire mitigation scenario and curriculum learning - a particular learning paradigm that relies on environment families.

Leveraging Multimodal Large Language Models (MLLMs) via contrastive learning has become a mainstream paradigm for improving the performance of Universal Multimodal Retrieval (UMR). However, previous works have ignored the grain blindness when adapting the contrastive paradigm into retrieval tasks. Grain blindness refers to the tendency of the model to overlook grain-level information contained in the query, which is crucial for effectively handling complex queries. This stems from contrastive learning treating samples as a binary classification (positive/negative), while ignoring the different information carried by each negative sample. To address this, we argue that negatives should be treated differently according to their similarity to the positive sample, enabling the model to learn distinct grain information from each negative. In this paper, we introduce a simple but effective framework, called ELVA, a novel rule-based RL framework that mitigates grain blindness through ranking-driven MLLMs. 1) Instead of relying on reward models, we extend Reinforcement Learning with Verifiable Rewards (RLVR) to retrieval tasks, allowing the model to explore new ranking behaviors without explicit ranking labels. 2) By utilizing rule-based rewards, our approach jointly optimizes the ranking of negative samples while enlarging the similarity gap between positive and negative. To more precisely measure grain blindness, we further introduce MRBench, a new benchmark specifically designed for multi-grain query scenarios. ELVA achieves state-of-the-art results across standard retrieval benchmarks, and its notable 13.1% improvement on MRBench further demonstrates its effectiveness in alleviating grain blindness.

Many decision-making problems in computing and networking systems can be naturally formulated as cost-minimization problems under performance constraints. In dynamic environments, reinforcement learning (RL) is often used to solve such problems at runtime by embedding both costs and constraint violations into a single scalar reward through weighted penalty terms, following a Lagrangian-inspired formulation. However, in this context the behavior of the learned policy critically depends on the choice of these weights, which are typically selected manually. This makes it difficult to identify an appropriate trade-off between optimizing the primary objective and effectively avoiding constraint violations, particularly in non-stationary environments where their relative importance may change. This paper presents MAMO (Multi-Agent system for Multi-Objective constrained optimization), an approach to tackle this balancing problem through multi-agent RL. MAMO decouples task execution from objective design by formulating the selection of reward weights as a learning problem, providing a !rst step towards more autonomous and robust RL-based solutions for constrained optimization problems in dynamic environments.

Immersion in video games depends not only on graphics, audio, and game mechanics, but also on the quality of in-game characters. Producing believable characters, or game AI, remains a significant challenge as behavioral complexity is hard to capture with hand-coded systems. Game AI is a source of immersion and engagement; however, the limitations stemming from the challenges of creating game AI often lead to frustration and the breaking of the illusion of realism within the game. The introduction of machine learning models opens the door to creating more believable, authentic, and relatable characters in games. The promise is that they either learn from interacting with the game, or from player data, to develop true human-like behavior. In this paper, we envision more applications of reinforcement learning for game AI in the future. For this to materialize, current research limitations are prohibitive to broad deployment across game genres. Therefore, we propose a framework for training reinforcement learning models with a set of requirements in mind that are suited towards game AI and game development. We present examples of games with reinforcement learning-augmented game AI and describe the practicalities of deploying player-facing machine learning agents in modern games. Furthermore, we identify bottlenecks and hard problems in these areas, which we believe offer promising research directions to accelerate the adoption of machine learning in game AI for the video game industry.

In offline Reinforcement Learning, immediate rewards in logged batch data are often unobserved due to sparse or irregular record-keeping, or censored beyond certain reward values. This issue arises in practical settings, including health care and marketing. We investigate off-policy evaluation (OPE) in finite-horizon Markov decision processes when rewards are missing not at random (MNAR), which breaks ignorability and induces selection bias even after conditioning on states and actions. To address this, we formalize a reward-dependent propensity model and use future states as shadow variables to identify the full-data conditional mean reward. We further introduce a bridge function that recovers the conditional mean reward without explicitly modeling the MNAR mechanism, and estimate it via a min-max procedure to avoid double sampling. Building upon these identification results, we propose an Fitted-Q-Evaluation-style estimator that propagates the recovered rewards while allowing target policies to depend on past missingness indicators. Finally, we establish consistency and finite-sample error bounds for our OPE estimator, and show through experiments the strong performance of our method compared to existing methods on simulated and MIMIC-III Sepsis data.

Optimal Reinforcement Learning (RL) algorithms typically rely on carefully constructed count-based uncertainty estimates to drive exploration. Although theoretically sound, such estimates are hard to compute in practical settings and therefore offer limited insight for designing exploration heuristics. Meanwhile, ensembling has emerged as a practical approach, but remains without theoretical justification. Building on a recent ensemble-based method for Multi-Armed Bandits, we propose a quantile-based ensemble method for finite-horizon Markov Decision Processes (MDPs). Our simple count-free approach achieves optimal variance-dependent regret bounds, providing theoretical grounding for ensemble-based exploration in RL.

While reinforcement learning from verifiable rewards (RLVR) typically has relied on a single binary verification signal, symbolic proof assistants in formal reasoning offer rich, fine-grained structured feedback. This gap between structured processes and unstructured rewards highlights the importance of feedback that is both dense and sound. In this work, we demonstrate that the Lean proof assistant itself can serve as a symbolic process oracle, supplying both outcome-level and fine-grained tactic-level verified feedback during training. Proof attempts are parsed into tactic sequences, and Lean's elaboration marks both locally sound steps and the earliest failing step, yielding dense, verifier-grounded credit signals rooted in type theory. We incorporate these structured rewards into a GRPO-style reinforcement learning objective with first-error propagation and first-token credit methods that balances outcome- and process-level advantages. Experiments with STP-Lean and DeepSeek-Prover-V1.5 show that tactic-level supervision outperforms outcome-only baselines in most settings, delivering improvements on benchmarks such as MiniF2F and ProofNet. Beyond empirical gains, our study highlights a broader perspective: symbolic proof assistants are not only verifiers at evaluation time, but can also act as process-level reward oracles during training. This opens a path toward reinforcement learning frameworks that combine the scalability of language models with the reliability of symbolic verification for formal reasoning.

Deep research agents have demonstrated remarkable capabilities in autonomous information gathering and synthesis, yet their training remains constrained by the static nature of simulated environments, the limits of fact-retrieval-only task designs, and the inefficiency of outcome-based reinforcement learning. In this work, we propose MetaResearcher, a novel framework that scales deep research agent training across four synergistic dimensions. First, we introduce an Evolving Virtual World that injects temporal dynamics and adversarial misinformation into the training environment, forcing agents to develop source credibility assessment and temporal conflict resolution skills. Second, we design Discovery-Oriented Tasks -- including hypothesis generation and contradiction resolution -- that transcend simple fact retrieval and push agents toward genuine research behaviors. Third, we propose a Self-Reflective Meta-Reward mechanism within the GRPO framework that jointly optimizes for answer correctness, search path efficiency, reflection depth, and tool call diversity, directly addressing the repetitive action loop problem observed in prior work. Fourth, we introduce a Heterogeneous Multi-Agent Swarm architecture comprising specialized Scout, Filter, and Synthesizer models that learn collaborative research strategies through coordinated reinforcement learning. Built upon the LiteResearcher infrastructure, MetaResearcher requires zero marginal API cost for training while targeting substantial improvements in both benchmark performance (GAIA, Xbench-DS) and epistemic robustness under adversarial conditions. We present the complete framework design, training methodology, and planned experimental validation.

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced Large Language Model (LLM) reasoning; however, it faces a fundamental optimization instability: uniform token updates precipitate entropy collapse, leading to premature convergence to suboptimal strategies, whereas excessive Shannon Entropy maximization can cause entropy explosion, driving blind exploration toward incoherent reasoning chains. To resolve this dichotomy, we introduce the Independent Combinatorial Tokens (ICT) framework, which shifts the optimization focus from scalar uncertainty to the distributional properties of token logits. By leveraging the Jensen-Shannon (JS) divergence between token logits distributions, ICT identifies tokens with distinctive distributional patterns as critical branching points for guiding effective exploration in LLM reasoning. Our theoretical analysis, grounded in both Shannon and second-order Rényi entropy, proves that selectively updating on these tokens regulates policy concentration: it reduces the overall distribution uncertainty measured by Shannon entropy, while controlling probability concentration captured by second-order Rényi entropy. This dual effect prevents over-concentrated token generation from weakening exploration and effectively stabilizes the training landscape. Empirical results demonstrate that updating only the top 10% of unique tokens on Qwen2.5 (0.5B/1.5B/7B) models yields an average pass@4 improvement of 4.58%, with a maximum gain of 14.9%, over GRPO, 20-Entropy, and STAPO baselines across seven benchmarks spanning math, commonsense, and Olympiad-level problems.

In challenging real-world reinforcement learning applications, communication delays or sensor failures often cause frame dropping, in which the agent cannot receive the dropped states and associated rewards. To address the performance degradation caused by frame dropping, the Decision Transformer under Random Frame Dropping (DeFog) was developed by incorporating additional mechanisms into the decision transformer to tackle frame dropping. Although DeFog can mitigate performance degradation in frame-dropping environments, since DeFog is an offline learning method, it struggles to effectively generalize to novel states not adequately represented in the training dataset. In this study, we propose OnDeFog, which integrates the mechanisms in DeFog with the online decision transformer (ODT), an online reinforcement learning method that learns policies through direct environmental interaction. Comprehensive experimental evaluation demonstrates that our proposed OnDeFog achieves superior performance compared to ODT in environments characterized by high dropping frame rate and outperforms DeFog on datasets containing a large amount of low-reward data.

From the past few years, web intelligent enhancement systems increasingly rely on heterogeneous and dynamic web data to deliver personalized, context-aware services. However, traditional machine learning, deep learning, and reinforcement learning models often struggle with semantic understanding, adaptability, and scalability in continuously evolving web environments. In this research, a Multi-Granular Attention-based Reinforcement Web Intelligent Enhancement System (MGAR-WIES) is proposed to address the challenges by integrating semantic graph modeling, attention mechanisms, and adaptive reinforcement learning. Initially, heterogeneous web data comprising structured, semi-structured and unstructured sources are collected and preprocessed for generating unified feature representations. These representations are transformed into a dynamic semantic graph, where entities and their relationships are modeled by using graph embeddings enhanced by attention mechanisms for capturing both local relevance and global contextual dependencies. Subsequently, an adaptive multi-agent reinforcement learning strategy leverages the attention-aware semantic states to optimize personalized web actions like content recommendation, navigation optimization, and service adaptation. Finally, the continuous online feedback is further integrated to update graph representations and learning policies in real time by ensuring sustained adaptability and performance. The proposed MGAR-WIES acheived better results in terms of accuracy (80%) when compared with existing approaches.

Self-improvement can self-regress. In REINFORCE post-training for code, a model can quickly improve on its optimized metric and then collapse within the same training campaign. We study this in a controlled multi-seed testbed using Qwen-2.5-3B and Qwen-2.5-7B, trained on competitive-programming tasks with binary CodeGrader reward across 10 sequential 20-step campaigns. Across campaigns, pass@1 shows a robust rise-then-collapse pattern: it peaks within tens of gradient steps and then falls back, sometimes to near zero. This is not cross-task catastrophic forgetting, but within-task policy over-optimization on a fixed distribution; KL- and EWC-style constraints do not prevent it. We ask where the control loop should sit. We compare three levels: CARE, a between-campaign memory mechanism with a capability posterior, transfer gate, and regression-aware belief revision; ES, a within-campaign early-stop rule that rolls forward the peak checkpoint and sets the next budget to peak_step+3; and GRPO, which changes the RL update using group-relative reward normalization. The answer is regime-dependent. On Qwen-2.5-3B, where naive REINFORCE is fragile, CARE v2 nearly doubles end-of-chain pass@1 from 4.9% to 9.5%, with paired bootstrap 95% CI [+0.4,+8.9] and gains in 4/5 seeds. On Qwen-2.5-7B, CARE reaches parity with naive REINFORCE, 13.8% vs. 11.8%, while ES reaches 22.2% [14.1,28.0]. Out-of-the-box GRPO reaches 20.7% [15.7,25.1], nearly matching REINFORCE+ES. GRPO raises the floor but does not remove the cliff. Its 7B gain mainly comes from better between-campaign carryover, while the within-campaign peak-to-end gap remains about 17 points under both REINFORCE and GRPO. GRPO+ES gives mixed evidence: 2/3 seeds improve, but one final cliff lowers the mean to 17.0% [0.0,28.1]. A Gemma-3-4B pilot shows the same signature, suggesting the phenomenon is not limited to Qwen.

Score- and flow-matching models often rely on preference-based reinforcement learning for two purposes: aligning with subjective preferences and, surprisingly, recovering properties such as visual realism and coherent object structure that matching-based training is intended to learn from the data itself. We argue that this reflects a structural mismatch. Matching losses measure $\ell_2$ regression error on the velocity or score field under training-time marginals, a proxy poorly aligned with the visual and semantic properties that determine sample quality at inference. Given a reward aligned with these properties, RL sidesteps the mismatch by evaluating the model on its own samples and following the reward landscape directly. The challenge is to obtain such a reward without relying on human preferences, which are expensive and conflate data realism with annotator inclinations. We propose Discriminator-Guided RL (DRL). DRL trains a discriminator to separate data from base-model samples in a pretrained representation space and uses its logit as the reward in KL-regularized RL. The pretrained space restricts the discriminator to perceptually meaningful directions, and the logit estimates the log-likelihood ratio between data and model, which is the optimal reward for targeting the data distribution. Across SiT, JiT, REPA, and RAE, DRL reduces guidance-free FID (e.g., $9.38 \to 2.62$ on SiT) and semantic-space FD (e.g., $88.2 \to 19.3$ on DINOv3 for SiT), with consistent gains across all backbones, and improves human-preference rewards without training on them. It also yields a better Pareto frontier between preference reward and image fidelity under subsequent preference-based post-training, increasing alignment while reducing low-level artifacts such as oversaturation and excessive brightness.

We present Pareto Q-Learning with Reward Machines (PQLRM), a multi-objective reinforcement learning algorithm for tasks whose reward structure is specified by a set of reward machines (RMs). PQLRM combines Pareto Q-Learning (PQL), which maintains sets of vector-valued Q-estimates to approximate the Pareto front, with enhancements from Q-Learning with Reward Machines (QRM), which exploits the factored automaton structure of the reward signal. This yields a multi-policy algorithm that remains sample-efficient under non-Markovian, RM-encoded rewards. Experimental trials show that PQLRM converges faster than a naive PQL baseline applied to the cross-product MDP and can synthesize Pareto-optimal policies that QRM cannot.

Reinforcement learning (RL) has become a representative post-training paradigm for large language models (LLMs), enabling strong reasoning and agentic capabilities. However, rollout generation remains a dominant latency bottleneck because autoregressive (AR) sampling decodes responses sequentially and a small number of long-tailed generations often determine completion time. Speculative decoding (SD) is a well-established technique for serving fixed LLMs that reduces latency by rapidly drafting tokens and accepting them through parallel verification while preserving the target-model distribution. However, its practical speedups do not directly carry over to RL rollouts: (i) the evolving target policy makes any fixed drafter increasingly mismatched with the policy's output distribution; and (ii) active batch sizes shrink throughout rollout decoding, shifting decoding from compute-bound to memory-bound regimes where parallel verification can exploit underutilized compute. Therefore, accelerating RL rollouts requires both a drafter that remains effective under long, high-temperature generations from an evolving policy and system-aware use of SD that avoids compute-bound regimes. We present EfficientRollout, a system-aware self-SD framework designed to address this gap for RL rollouts. EfficientRollout induces a quantized drafter from the target model (i.e. self-speculative decoding), keeping it coupled to the evolving policy without separate drafter pretraining or online adaptation. It further coordinates a system-aware SD toggle policy with acceptance-aware draft-length adaptation, enabling speculation only in beneficial regimes while matching the drafting budget to evolving drafter quality. EfficientRollout reduces rollout and end-to-end latency by up to 24.2% and 15.6%, respectively, over an accelerated AR rollout baseline, while preserving final model quality.

We study online reward-punishment learning when the environment provides no scalar reward or evaluative label. At each step the agent receives only a fixed-channel perceptual packet, and quantities such as pain, energy, contact, damage, or cognitive error are treated as perceptual dimensions whose valence must be inferred from transition consequences. OHIRL separates four roles: M_psi learns next-packet prediction, D_omega models residual dynamics, C_eta is a fixed internal post-transition trajectory evaluator, and B_xi learns to use the resulting value evidence for later policy updates and action scoring. C_eta uses a recovery-positive and persistence/growth-negative residual-regulation orientation; a coefficient-origin audit shows that equal-unit, raw-equal, and random monotone variants preserve more than 92% of the released top-action rankings, while sign inversion preserves 0%. The reward-free protocol exposes observation transitions while withholding environment rewards, delayed external evaluators, success labels, and action-goodness labels. A conditional error decomposition separates B_xi evidence-estimation error from residual policy-optimization error. In a 2x2-XOR packet task, medicine and chili acquire opposite value under visual XOR contexts, and the same pain or spice increase can be positive or negative depending on consequence structure; B_xi reaches 0.952 balanced reward-sign accuracy. In a full online-interleaved audit, M_psi reaches holdout R2=0.907, B_xi reaches 0.940 sign accuracy, and the policy reaches 0.979 optimal-action accuracy, while immediate packet scores, prediction-error rewards, shuffled targets, zero reward, and error-reduction controls collapse. Hidden-reward CartPole and Taxi controls, public-context no-leakage audits, and module-role ablations further test information boundaries and component necessity.

Protein language models (PLMs) have emerged as powerful tools for controllable biomolecular design, yet their post-training adaptation typically relies on costly wet-lab validation or curated preference datasets. To overcome this supervision bottleneck, we introduce unsupervised reward optimization of PLMs, a comprehensive framework for steerable protein generation without ground-truth labels. Our key insight is that task-agnostic rewards, which combine intrinsic model uncertainty with extrinsic semantic consistency informed by protein representation models, exhibit strong correlation with controllability measures across base models and temperature regimes. Building upon this discovery, we propose two offline algorithms: Soft Reward Optimization (SRO) and Binarized Reward Optimization (BRO), which effectively maximize the classical RLHF objective induced by these proxy rewards. Extensive experiments on compositional out-of-distribution prompts demonstrate that both methods significantly outperform competitive baselines (DPO, KTO), while approaching oracle performance across multiple sampling temperatures, model scales and protein families. Moreover, PLMs fine-tuned with unsupervised rewards can achieve consistently higher coverage compared to their base model in pass@k evaluations. By enabling self-improvement of PLMs through their own generated experience, our framework provides a scalable pathway toward controllable biomolecular design in settings where labeled preferences or experimental feedback are scarce or unavailable.

Reinforcement Learning with Verifiable Rewards (RLVR) has become a standard paradigm for enhancing the capability of large reasoning models. RLVR typically samples responses independently and optimizes the policy using from final answers. This paradigm has two limitations. First, independently responses often contain similar intermediate reasoning steps, causing redundant exploration and wasted computation. Second, sparse final-answer rewards make it hard to identify useful steps. Tree-based methods partly address this problem by sharing prefixes and comparing branches from the same prefix to provide fine-grained signals. However, tree branches are still expanded independently. When different branches reach similar reasoning states, they cannot share information and repeat similar exploration. Moreover, tree-based methods ignore such dispersion and only perform local comparisons within separate branches, which can lead to higher variance in advantage estimation. To address this challenge, we propose GraphPO (Graph-based Policy Optimization), a novel RL framework that represents rollouts as a directed acyclic graph, with reasoning steps as edges and semantic states summarized from the reasoning paths as nodes. GraphPO merges semantically equivalent reasoning paths into equivalence classes, allowing them to share suffixes and reallocating budget away from redundant expansions to diverse exploration. Furthermore, we assign efficiency advantages to incoming edges and correctness advantages to outgoing edges, thereby improving inference efficiency while deriving process supervision from outcome. Theory shows that GraphPO reduces advantage-estimation variance and enhances reasoning efficiency. Experiments on three LLMs across reasoning and agentic search benchmarks show that GraphPO consistently outperforms chain- and tree-based baselines with the same token budgets or response budgets.

Vision-Language-Action (VLA) models can generalize across diverse manipulation tasks, but their imitation-learning-based policies remain brittle in precise physical interactions due to compounding execution errors; Can a reinforcement learning policy trained purely in simulation improve the robustness of real-world VLAs zero-shot? Residual RL, which learns a corrective policy on top of a frozen VLA, offers a natural framework, but existing approaches face a fundamental sim-to-real dilemma: privileged-state methods require lossy distillation for deployment; image-based methods suffer from the visual domain gap; and real-world RL is costly and unsafe. We propose an object-centric residual RL framework that refines VLA actions using object poses, enabling a compact observation space that transfers consistently between simulation and reality. To align the two domains, we additionally replay the same teleoperation demonstrations in simulation to train a sim counterpart of the real-world VLA. The residual RL policy is trained only in simulation with pose noise injection and dropout, and transfers zero-shot to the real robot. Across five manipulation tasks on a real Franka Research 3 (FR3) robot, our method improves the success rate from 42% to 76% zero-shot, and the improved rollouts can be further reused to retrain the base VLA for self-improvement without additional teleoperation. Project page: https://www.microsoft.com/en-us/research/articles/object-centric-residual-rl/

Foundation models for language and vision are powered by internet-scale data, while structured domains such as tabular prediction are powered by synthetic data. This substitute shifts the challenge from collection to prior design. Such priors already exist for many structured tasks: TabPFN and its successors solve tabular classification with a transformer pretrained on a synthetic Bayesian prior. We make two points. \textbf{First}, reinforcement learning is the conspicuous gap: sampling a synthetic MDP is as feasible as sampling a synthetic tabular dataset, yet no in-context RL work treats prior design as a primary objective. \textbf{Second}, MDPs admit a fixed-size sufficient statistic, independent of the episodes observed and tabular in shape, which makes them directly amenable to the attention-based architectures used for tabular foundation models, with a policy head replacing the supervised target. Together these define the agenda for an RL foundation model. As a proof of concept, we train a Graph Attention Network entirely on synthetic MDPs and show that, with no task-specific tuning, it solves held-out tabular benchmarks in context, both online and offline: online, in far fewer episodes than UCB-VI and tabular Q-learning, and offline, competitively with VI-LCB.

In real-world reinforcement learning (RL), the choice of action space can play a key role in shaping motion smoothness, safety, and overall task performance. In this study, we evaluate pose increment, pose velocity, joint position increment, and joint velocity across two vision-based manipulation tasks: object picking and pushing. We train policies in simulation and deploy them to the real world using sim-to-real transfer. We find that action-space representation indeed significantly affects sim-to-real performance. In particular, we find that the joint velocity action space is best for the vision-based picking and pushing tasks in terms of smoothness and final task performance. We also provide practical guidance for RL practitioners in choosing action spaces for both simulation and real-world experiments.

Humans often acquire new skills by observing others, since observed behaviors implicitly reveal how to act in an environment. However, observations drawn from a heterogeneous population introduce conflicting behavioral signals, making it difficult to determine which behaviors are worth imitating. We address this challenge with General Reward Inference and Disentanglement (GRID), a social learning method that extracts universally useful behaviors from a heterogeneous population of demonstrators pursuing different goals. GRID decomposes per-agent reward functions into a general reward, capturing behaviors shared across all agents, and specific rewards, capturing individual preferences and objectives. Training exclusively on the general reward provides a new paradigm of generalist pretraining. It yields a generalist agent that internalizes universal environmental competencies, such as safety and basic task proficiency, without the mode-averaging bias that afflicts standard learning from demonstration techniques. This generalist serves as a superior prior for fine-tuning to downstream tasks, including preferences unseen during training. Experiments across a synthetic basis function decomposition, multi-agent Craftax, and a continuous autonomous driving simulator (Highway-Env) confirm that GRID successfully disentangles reward structure in a semantically meaningful way, outperforms standard learning from demonstration baselines, and enables more efficient and stable specialization.

Multi-fuel compression-ignition engines offer fuel flexibility but introduce uncertain, time-varying fuel reactivity, represented by cetane number (CN), which complicates cycle-to-cycle combustion-phasing control. This work formulates CA50 regulation under latent CN variation as a partially observable sequential decision problem and systematically evaluates controllers with increasing temporal and representational capacity, including LinUCB, history-augmented contextual bandits, observation-only DDPG, recurrent DDPG, and a proposed GRU-guided RL framework. A Gaussian-process surrogate trained on experimental multi-fuel engine data provides a controlled and reproducible evaluation environment. Results show that myopic and fixed-history bandit methods degrade under CN variation, observation-only RL suffers from latent-state aliasing, and generic recurrence is insufficient when CN evolves rapidly. The proposed framework learns a compact GRU-based representation of fuel reactivity from combustion history and conditions both actor and critic on this estimated signal rather than oracle CN. By training the policy on the same imperfect fuel-reactivity information available at deployment, the controller avoids train-deploy inconsistency in conventional online estimate-then-control pipelines. Across unseen CN trajectories, the policy achieves stable CA50 regulation with mean absolute tracking error below 0.25° CA at the training setpoint, while producing smooth, physically consistent SOI and glow-plug-power actuation. These results show that combustion control under latent, continuously evolving fuel dynamics requires more than standalone estimation or generic recurrence. By aligning fuel-reactivity inference with control policy learning, the proposed framework enables reactivity-aware decision-making using the same estimated state available during deployment.

RL post-training strategies are dataset-dependent and reveal a recurring empirical pattern: capacity parameters accumulate monotonically across stages, while regularization parameters predominantly oscillate in response to shifting training dynamics. This distinction highlights a potential flaw in fixed training schedules: by forcing all parameters along rigid paths, they fail to capture the dynamic exploration-exploitation tradeoffs that regularization must track. We uncover this through LLMZero, an agentic system that optimizes training trajectories via tree search by diagnosing pathologies at each checkpoint and proposing coordinated multi-parameter transitions. Across four diverse GRPO tasks, LLMZero discovers strategies that improve over the base model by 9% to 140% and over grid search by 6% to 15% (relative), consistently outperforming random search and a skill-based agent under a matched compute budget. The capacity--regularization asymmetry is consistent across all four tasks, offering a candidate design heuristic for multi-stage training.

This paper explores the problem of finding the minimum zero-forcing set on undirected graphs and proposes an adapted machine-learning framework to solve the problem. The minimum zero-forcing set problem is a graph coloring problem where the color of an initial set of nodes propagates throughout a network. The set of nodes is zero-forcing if it forces all uncolored nodes to change color under the constraint of the color-change rule. There are several applications to this problem across different domains such as network science, network control, and designing logical circuits. Finding the minimum zero-forcing set is shown to be NP-hard. We propose a reinforcement learning framework, SD-ZFS, that adapts the S2V-DQN architecture to the ZFS problem. We train several models on this adapted framework and analyze the performance across graph datasets that have varying structures. We evaluate how the models trained on the framework generalize, scale, and transfer to different network types. The results demonstrate the effectiveness of the framework when compared against the optimal solution and greedy heuristic. We provide further insight into how the ZFS problem can be solved through machine-learning and the influence of network structure on the problem.