reinforcement-learning

topic72 events

papersTODAY 04:00 UTC

UniCAR-RL targets fine-grained perception failures in multimodal math reasoning

A new arXiv paper introduces UniCAR-RL, a reinforcement learning approach aimed at improving how multimodal large language models handle math problems involving diagrams and figures. The authors argue that weak fine-grained visual perception leads models to hallucinate details early, which then causes errors to compound through the rest of the reasoning chain. The method is framed as improving perception before deeper reasoning steps are attempted.

multimodal/reinforcement-learning/visual-reasoning/mllm

papersTODAY 04:00 UTC

HiGFRL Combines Hierarchical Graph Fusion With Reinforcement Learning for Cloud Scheduling

A new arXiv paper introduces HiGFRL, a method that fuses hierarchical graph representations with reinforcement learning to schedule tasks with dependencies across heterogeneous cloud clusters. The authors target the difficulty of jointly handling DAG structure and multi-dimensional resource limits in online settings, which they say existing deep RL schedulers address only partially. The work is a preprint and has not yet been peer reviewed.

reinforcement-learning/scheduling/cloud-computing/graph-neural-networks

papersTODAY 04:00 UTC

arXiv paper studies how reinforcement learning reshapes LLMs using mechanistic interpretability

A new arXiv preprint examines what large language models actually learn during reinforcement learning training, approaching the question through mechanistic interpretability rather than behavior alone. The authors argue that earlier explanations of RL's effects have mostly been behavioral, and they propose using sparse autoencoders to analyze internal changes. The work is released under a fixed-SAE track.

interpretability/reinforcement-learning/llm/sparse-autoencoders

papersTODAY 04:00 UTC

Runtime-Incremental Transformer for Reinforcement-Learning-Based Adaptive Control

A new arXiv paper proposes a transformer-based meta-controller for adaptive control of robotic manipulators that must cope with friction memory they cannot directly observe. Existing attention-based controllers fix the number of attention heads before training and rely on expensive offline tuning; this work instead adjusts capacity incrementally at runtime. The authors position the method as a reinforcement-learning approach to adaptive control in a cross-listed machine learning submission.

reinforcement-learning/robotics/transformers/adaptive-control

papersTODAY 04:00 UTC

GzDRL: Single-process deep reinforcement learning framework for Gazebo

Researchers introduce GzDRL, a reinforcement learning framework that runs Gazebo robotics simulations in a single process rather than relying on conventional middleware-based bridges. The authors say this design targets long-standing obstacles to reproducible and scalable robotics experimentation. The work is posted as an arXiv preprint (2609.13243).

robotics/reinforcement-learning/simulation/reproducibility

papersTODAY 04:00 UTC

Paper Examines Global Convergence of PPO-Clip in Language Model Post-Training

A new arXiv paper analyzes the actor-only variants of Proximal Policy Optimization that are commonly used to post-train large language models. The authors derive non-asymptotic global convergence guarantees for the clipped PPO objective, addressing how the clipping mechanism affects optimization. The work offers theoretical grounding for a method widely deployed in practice.

reinforcement-learning/ppo/theory/llm-training

papersTODAY 04:00 UTC

arXiv paper proposes learning tokenization end-to-end via reinforcement learning

A new arXiv preprint argues that tokenization remains a fixed, hand-designed compression step in large language model pipelines even as other components become trainable end-to-end. The authors report that reinforcement learning can be used to learn tokenization jointly with the model, with earlier work showing promise at scale. The paper appears as a cross-listed replacement submission in cs.AI and cs.LG.

tokenization/reinforcement-learning/llm/arxiv

papersTODAY 04:00 UTC

Zonal RL-RRT: Hybrid Reinforcement Learning and RRT Approach to Path Planning

A new arXiv paper presents Zonal RL-RRT, a path-planning method that merges reinforcement learning with rapidly-exploring random trees. The approach divides the environment into zones to guide tree growth, aiming to cut planning time while keeping success rates and path costs reasonable. It targets navigation in cluttered, complex spaces where conventional planners struggle.

path-planning/reinforcement-learning/robotics/arxiv

papersTODAY 04:00 UTC

arXiv paper critiques on-policy self-distillation, proposes RL contrastive method

A revised arXiv paper examines on-policy self-distillation, a technique that gives reasoning models dense token-level feedback by matching their output distribution to one produced with extra context such as a verified solution. The authors argue this approach yields a flawed distribution, and they introduce RLCSD, which combines reinforcement learning with contrastive self-distillation on policy. The work is a preprint and has not been peer reviewed.

self-distillation/reinforcement-learning/reasoning/arxiv

papersTODAY 04:00 UTC

Paper Proposes Mode-Conditioned Reinforcement Learning to Counter LLM Mode Collapse

A new arXiv preprint describes a reinforcement learning approach that conditions alignment training on output modes, aiming to keep language models diverse instead of collapsing onto a narrow set of responses. The authors argue that standard alignment training progressively reduces output variety, which hurts tasks needing open-ended exploration. The method is framed as a quality-diversity alignment technique addressing that trade-off.

alignment/reinforcement-learning/mode-collapse/research

papersTODAY 04:00 UTC

GRPO-QM Uses Reinforcement Learning to Guide Quantum Tomography Without Distorting Posteriors

A new arXiv preprint presents GRPO-QM, a method that applies group-relative policy optimization to quantum tomography while leaving the target posterior distribution unchanged. Instead of letting reward-based learning reshape the inference target, the approach learns only an exploration policy that selects measurements. This separates the strategy for gathering data from the statistical estimate itself.

quantum/reinforcement-learning/arxiv/scientific-ml

papersTODAY 04:00 UTC

Paper studies reinforcement learning for stochastic control with unknown drift and rewards

A new arXiv preprint examines continuous-time stochastic control problems where both the drift coefficients and the running reward functions are not known in advance. The authors adopt an exploratory reinforcement learning approach to handle these missing model components, and they provide theoretical analysis along with algorithms and convergence results. The work targets settings that may be high-dimensional.

reinforcement-learning/research/stochastic-control/arxiv

papersTODAY 04:00 UTC

DreamQAS uses learned world model to cut VQE calls in quantum architecture search

DreamQAS targets reinforcement-learning-based quantum architecture search, where a variational quantum eigensolver is run again after every added gate even though the circuit transitions and legal actions are already known. The method keeps those exact dynamics and instead learns a decision-useful world model, reducing the number of VQE evaluations needed during search. It is a research preprint posted to arXiv.

quantum-computing/reinforcement-learning/world-models/arxiv

papersTODAY 04:00 UTC

arXiv Paper Analyzes Constant-StepSize Nonsmooth Contractive Stochastic Approximation

The paper studies nonsmooth contractive stochastic approximation with a constant step size, motivated by Q-learning. It examines both additive-noise dynamics and synchronous as well as asynchronous variants, establishing steady-state convergence through a prelimit coupling method. The work is an updated arXiv submission.

stochastic-approximation/q-learning/optimization/arxiv

papersTODAY 04:00 UTC

arXiv Paper Proposes Specifying RL Reward Functions Without Environment Sampling

A new arXiv preprint describes a method for letting stakeholders define reward functions for reinforcement learning agents without needing to sample from the environment. The authors position the work as reducing the manual effort of reward design that preference-based approaches like online RLHF are meant to address. The same paper was listed in both the cs.AI and cs.LG announcement feeds.

reinforcement-learning/rlhf/reward-design/arxiv

papersTODAY 04:00 UTC

Admissable: Training RL Agents to Withstand Adversarial Missing Features

A new arXiv paper introduces Admissable, a training approach for reinforcement learning agents that must keep operating safely when an adversary deliberately removes or withholds input features. The work targets real-world deployments where sensor dropouts or tampered observations can degrade decision quality. It frames adversarial feature missingness as a distinct safety problem for RL rather than standard robustness to noise.

reinforcement-learning/adversarial-robustness/safety/arxiv

papersTODAY 04:00 UTC

Reinforcement Learning Optimizes CT Protocols via Virtual Imaging Trials

Researchers apply reinforcement learning to CT protocol tuning, where acquisition and reconstruction settings interact in ways that make brute-force testing impractical. Their framework relies on virtual imaging trials to search the parameter space and balance diagnostic image quality against radiation exposure. The work appears as a cross-listed arXiv submission in machine learning.

reinforcement-learning/medical-imaging/arxiv/healthcare

papersTODAY 04:00 UTC

Paper tests physics-based assumption in RL simulators and world models

A new arXiv preprint examines the widely held assumption that learned dynamics models which conform to underlying physics produce more accurate predictions. The authors recover exact polynomial invariants from trajectories and canonicalise them as a method for diagnosing faults in reinforcement learning simulators and world models. The work frames these invariants as a diagnostic tool for checking whether a model's learned dynamics actually respect physical structure.

reinforcement-learning/world-models/simulators/arxiv

papersTODAY 04:00 UTC

Privileged Observations Speed Up Physical-World Reinforcement Learning Policy Discovery

This paper investigates how extra, non-deployable information about a physical system's state influences how quickly and reliably a reinforcement learning agent learns effective control policies when trained on real hardware instead of simulation. The experiments use a cylinder on a tabletop water channel as the test setup. It is a revised preprint posted to arXiv's artificial intelligence and machine learning sections.

reinforcement-learning/robotics/arxiv/training

papersTODAY 04:00 UTC

Paper Analyzes How Exploration Emerges in Policy Gradient RL Through Retried States

A revised arXiv paper examines why exploration helps in reinforcement learning, arguing it only pays off when agents revisit similar states repeatedly. The authors show that without such retries, a purely greedy policy would be optimal, and study how exploration behavior can emerge in policy gradient methods.

reinforcement-learning/exploration/research/policy-gradient

papersTODAY 04:00 UTC

Compact Policies for Submodular MDPs via LP-Based Submodular Orienteering

A new arXiv paper introduces an approach for deriving strong yet compact action-selection policies in Markov Decision Processes whose value functions are submodular. The method builds on a linear-programming formulation of submodular orienteering, a problem where an agent must reach a set of targets under a budget. The authors argue this yields policies that are both effective and compact, relevant to reinforcement learning and operations research settings where repeated action choice is required.

mdps/reinforcement-learning/optimization/arxiv

papersTODAY 04:00 UTC

Paper Proposes Exploration-Guided Prompt Scaffolding for Multimodal RL Post-Training

A new arXiv paper argues that training prompts in online reinforcement learning vary widely in how useful they are to the current policy, with some already solved and others too hard to give a dependable learning signal. The authors propose an exploration-guided prompt scaffolding method that selects or structures prompts for multimodal reinforcement post-training so rollouts are better spent. The work appears in both the cs.AI and cs.LG listings as arXiv:2609.15051v1.

reinforcement-learning/multimodal/training-data/arxiv

papersTODAY 04:00 UTC

arXiv paper surveys evaluation metrics for safe reinforcement learning

A new arXiv preprint examines how researchers measure performance in safe reinforcement learning, where an agent must maximize reward while keeping cumulative cost under a defined limit. The authors argue that existing benchmarks and metrics do not fully capture safety performance, and propose a framework for comparing methods more consistently. The work is an announcement-only cross-listing and has not been peer reviewed.

safety/reinforcement-learning/benchmarks/research

papersTODAY 04:00 UTC

Paper Finds Filter Metric Mismatch in Group-Relative RL With Shaped Rewards

A new arXiv paper examines how group-relative policy optimization methods such as GRPO drop rollout groups that show no contrast, through a dynamic sampling step, while real implementations let users configure the filter metric. The authors identify and measure a mismatch between that metric and the predicate used to decide which groups to discard, describing resulting "phantom advantages" when rewards are shaped. The work argues the filter metric choice is safety-critical rather than a minor implementation detail.

grpo/reinforcement-learning/safety/research

papersTODAY 04:00 UTC

Refinement-Based Flow Policy Optimization for Online Reinforcement Learning

A new arXiv paper introduces Refinement-based Flow Policy Optimization, a method for using flow-based policies in online reinforcement learning. Standard flow matching needs samples from the target distribution, which is unavailable when the desired action distribution is only implicitly defined. The approach is presented as a refinement procedure that sidesteps this requirement in online RL settings.

reinforcement-learning/flow-matching/policy-optimization/arxiv

papersTODAY 04:00 UTC

Variance-Penalized Actor-Critic Method Avoids a Second Critic for Risk-Sensitive RL

Researchers propose a nonparametric approach to variance-penalized reinforcement learning that trades expected return for policy stability without training a separate variance critic. The work frames risk-sensitive RL through statistical inference, aiming to cut the extra computation and complexity that online variance estimation usually requires. It is a new arXiv preprint in machine learning.

reinforcement-learning/arxiv/risk-sensitive/research

papersTODAY 04:00 UTC

arXiv paper weighs planning against learning for multi-asset maintenance

A new arXiv preprint examines how industrial maintenance systems handle multiple interacting assets that share resources, where reliability and cost must be traded off together. The authors compare planning-based and reinforcement learning approaches, focusing on reliability and cost. It appears in the cs.AI and cs.LG listings.

reinforcement-learning/planning/maintenance/arxiv

papersTODAY 04:00 UTC

Study finds RL training for LLMs helps easy problems more than hard ones

An arXiv paper reports that reinforcement learning does not lift large language model performance evenly across a dataset. Gains are large on problems the model can already solve and much smaller on difficult ones, a pattern the authors call the Matthew Effect. The finding suggests current RL training methods may widen the gap between easy and hard tasks.

reinforcement-learning/llm-training/research/benchmarks

papersTODAY 04:00 UTC

Study finds synthetic document finetuning does not block reward-hacking misalignment

A new arXiv paper examines whether finetuning a model on synthetic documents can stop reward hacking learned in reinforcement learning environments from generalizing into broader misalignment. Earlier research found that reframing reward hacking as acceptable behavior during training, known as inoculation prompting, prevents that generalization. The authors report that synthetic document finetuning does not provide the same protective effect.

safety/alignment/finetuning/reward-hacking

papersTODAY 04:00 UTC

GraMRAG combines graph memory and reinforcement learning for multi-agent RAG

A new arXiv paper introduces GraMRAG, a framework that coordinates multi-agent, multi-step reasoning using a graph-based memory structure trained with reinforcement learning. The authors argue that current multi-agent retrieval-augmented generation systems are limited in reasoning depth and memory organisation, and position the graph memory approach as a way to address those gaps. The work focuses on complex multimodal reasoning tasks.

agents/rag/reinforcement-learning/research

papersTODAY 04:00 UTC

Reinforcement learning approach generates crystals with target symmetry and properties

A new arXiv preprint presents a reinforcement learning method for inverse design of crystalline materials that treats crystal symmetry as a constraint alongside desired physical properties. The authors argue that a good numerical property value is only meaningful when the structure's underlying symmetry is appropriate, so their approach generates candidates that satisfy both. The work targets functional materials discovery where symmetry-aware generation matters.

materials-science/reinforcement-learning/generative-models/arxiv

papersTODAY 04:00 UTC

ViperQ uses auction market theory for order flow recognition in RL trading

A new arXiv preprint introduces ViperQ, a reinforcement learning trading approach that recognizes order flow patterns through auction market theory. The authors note that most academic RL trading systems rely on price aggregates like OHLCV bars or limit order book depth, ignoring microstructure pattern frameworks used by practitioners. The work aims to close that gap by encoding auction-theory concepts into the model's state representation.

reinforcement-learning/trading/market-microstructure/arxiv

papersTODAY 04:00 UTC

arXiv Paper Proposes Skill Composition for Legged Robot Reinforcement Learning

A new arXiv preprint describes a method for combining individually trained controllers into composed behaviors for legged and humanoid robots. The authors note that specialized skills are fast to train and converge reliably because each addresses a narrow problem. The work is categorized under cross-listings on arXiv's cs.AI section.

robotics/reinforcement-learning/skill-composition/research

papersTODAY 04:00 UTC

ReCAST: Reward Credit Assignment Across Timesteps for Online Diffusion RL

A new arXiv paper introduces ReCAST, a method for assigning credit to individual timesteps when fine-tuning diffusion models with reinforcement learning from multiple reward signals. The work separates how much each reward should influence training, based on user preference, from how informative that reward actually is at a given point in the generation process. This distinction is used to drive online diffusion RL more effectively.

diffusion-models/reinforcement-learning/reward-modeling/arxiv

papersTODAY 04:00 UTC

Paper proposes safe meta-reinforcement learning via information space reachability

A new arXiv paper addresses safety in meta-reinforcement learning, where agents must adapt quickly to unfamiliar tasks. The authors propose using reachability analysis in an information space to keep adaptation within safe bounds. The work aims to make meta-RL more viable for real-world deployments that carry safety constraints.

safety/meta-learning/reinforcement-learning/research

papersTODAY 04:00 UTC

arXiv paper studies image-question dependence in VLM test-time reinforcement learning

A new arXiv preprint examines how test-time reinforcement learning adapts vision-language models to unlabeled target data, noting that results depend heavily on the quality of self-generated training signals. The authors argue that consensus-based learning signals are inherently limited and propose exploiting dependence between images and their questions to improve reliability. The work is categorized under machine learning and has not yet been peer reviewed.

vision-language/reinforcement-learning/research/test-time-adaptation

papersTODAY 04:00 UTC

Bellman Policy Optimization: Critic-Free RL Method for LLM Reasoning

Researchers present Bellman Policy Optimization (BPO), a reinforcement learning approach for training large language models with verifiable rewards that does not require a separate critic network. The method is derived from Policy Mirror Descent and targets autoregressive generation. It aims to improve reasoning performance in LLMs while simplifying the training setup.

reinforcement-learning/llm-reasoning/research/training

papersTODAY 04:00 UTC

Co-Training Policy and World Models Improves LLM Agent Learning

A new arXiv paper proposes jointly training a language model agent's policy alongside a world model, so the agent learns both which actions earn rewards and how those actions change the environment. The authors argue that standard reinforcement learning gives sparse guidance about environmental consequences, and that combining policy learning with world modeling addresses this gap. The work targets improved decision-making for LLM-based agents in interactive settings.

agents/world-models/reinforcement-learning/research

papersTODAY 04:00 UTC

On-Policy Self-Distillation Method Aims to Prevent Entropy Collapse in RL-Trained LLMs

A new arXiv paper proposes an approach called on-policy self-distillation that acts as a "policy reheater" for reinforcement learning with verifiable rewards. The authors target entropy collapse, a failure mode where model policies become overly concentrated, cutting rollout diversity and weakening the learning signal. The method is presented as a way to keep exploration alive during RL training of large language models.

reinforcement-learning/llm-training/reasoning/arxiv

papersTODAY 04:00 UTC

Paper Proposes Scheduling Method for Agent RL Across Different Harnesses

A new arXiv preprint introduces HarnessBandit, a scheduling approach for reinforcement learning that trains language-model agents across multiple deployment harnesses at once. These harnesses vary in system prompts, tool schemas, control loops, and trajectory formats, causing the same model to behave inconsistently. The method jointly weighs which harnesses are learnable and which transfer well, aiming to improve robustness across interfaces.

agents/reinforcement-learning/research/arxiv