Reinforcement learning
139 notes, newest first. Includes the vault tags reinforcement learning, policy optimization, policy learning, rlhf.
Homomorphic Advantage Operator: Stabilizing Reinforcement Learning Under Fully Homomorphic Encryption Constraints
The preprint proposes the Homomorphic Advantage Operator to stabilize deep reinforcement learning when nonlinear operations are approximated under fully homomorphic encryption. It applies a zero-mean centering…
PivotOPD: Learning to Recover from Pivotal Mistakes in Multi-Turn Agents
The technical report studies pivotal mistakes that move an agent farther from task completion during multi-turn interaction. Across three Qwen3 models, the abstract reports that more than half of failed rollouts…
Reinforcing Agentic Creativity in Scientific Ideation with Night Science
The preprint introduces AI Night-Scientist, an agentic framework that uses reinforcement learning to shape when and how a model departs from predictable reasoning. It models creativity along action, process, and…
Up and Down the Abstraction Ladder: Code-Based Skills for Language Agents
The preprint studies reusable code-based skills for agents that otherwise must repeatedly choose low-level actions. In NetHack, the authors compare primitive actions, semantic skills, and combinations of both across…
AD-WM: Action-Discriminative World Models for Counterfactual Model Predictive Control
AD-WM is a joint-embedding world model designed for counterfactual model predictive control. It combines residual latent dynamics with action-recovery regularization so the model preserves action-dependent differences…
SciWalker: Synthesizing Scientific Coding Problems with Operator Graphs and Execution Feedback
The preprint describes a framework for synthesizing scientific coding problems by sampling operator chains from graphs built from scientific library interfaces and operation modes. An LLM generates problem statements…
CodeMidas: Scaling Agentic Coding RL Environments from Code Itself
CodeMidas converts implemented functionality in open-source codebases into executable reinforcement-learning environments using source code as the only task-specific input. Agents formulate behavioral specifications…
Score Centering Stabilizes Off-policy Reinforcement Learning
The preprint attributes instability in large-language-model reinforcement learning under training-inference mismatch to persistent drift between engines. It derives an additive score-centering correction that cancels…
Bypassing inference bottlenecks: Accelerating complex AI search with Retrieve-for-Train
Google Research describes Retrieve-for-Train as an offline reinforcement-learning process that compiles set-level rewards into training data for a lightweight diffusion retriever. The associated arXiv paper describes…
EnvHarness: Awakening Static Worlds for Agent Learning
The paper proposes a new system called Environment Harness (EnvHarness) to improve static environments used in learning by AI agents. The system can adapt the behavior of these environments without modifying their…
EnvHarness: Awakening Static Worlds for Agent Learning
The authors propose a method called Environment Harness (EnvHarness) to address the issue of static environments in learning agents. This system allows for reshaping behavior without modifying the underlying logic…
SAPO: Single-Rollout Autoregressive Policy Optimization for Agentic Reinforcement Learning
A new reinforcement learning method called Single-rollout Autoregressive Policy Optimization (SAPO) is proposed to address the memory overhead, computation efficiency, and performance issues of existing methods. SAPO…
MLREF: Efficient Module Reuse for Reward Design in Reinforcement Learning via Large Language Models
The research proposes a new framework called Module Level Reward Evolution Framework (MLREF) to address the issue of unreliable reward function preservation and reuse in reinforcement learning. MLREF uses a module…
Chronocooked: A Benchmark for Implicit Interval Timing in Reinforcement Learning Agents
The paper introduces Chronocooked, a benchmark suite for studying how reinforcement learning (RL) agents make decisions over time. It's inspired by Overcooked and includes cooking scenarios that require temporal…
ClawGym II: Exploring Black-Box RL on Agent Harness
Researchers present a new framework for reinforcement learning (RL) using complex agent harnesses. This framework is designed to optimize RL more stably and efficiently for long-horizon tasks. The researchers' method…
UniRL: A Framework for Unified Multimodal Model Reinforcement Learning
Tencent Hunyuan released UniRL, an open-source framework that applies one shared reinforcement-learning training loop across different kinds of generative models (diffusion/image models, text language models…
Custom reward functions for multi-turn reinforcement learning with Amazon Nova Forge
AWS engineers describe how to design reward functions for training AI models on multi-turn tasks (e.g., a coding assistant that should ask clarifying questions before writing code) using Amazon Nova Forge, and how to…
Learning Loco-Manipulation From SMPC Demonstrations With Sparse Offline-to-Online RL
The paper proposes a method to integrate robot locomotion and manipulation using Reinforcement Learning (RL) and Sample-based Model Predictive Control (SMPC). The approach generates offline datasets from SMPC…
One Frozen Simulator Is Not Enough: Simulator Collapse in Multi-Agent RL
The paper discusses a common issue in multi-agent reinforcement learning for human-AI interaction: using a single large language model (LLM) to simulate user behavior often results in the LLM focusing on specific…
Redistribution-based Cost Inference Improves Sparse Safe Offline RL
A new framework called Redistribution-based Cost Inference (RCI) is proposed to convert sparse stop-feedback in safe offline reinforcement learning into dense per-step costs. This conversion helps train a constrained…
MIRA: Medical Image Reflection for Agentic Diagnosis
Researchers present MIRA, a medical image diagnosis framework that allows an autonomous agent to search for relevant evidence, verify the relevance and consistency of the acquired evidence, and adapt its tool-use…
MIRA: Medical Image Reflection for Agentic Diagnosis
The paper introduces MIRA (Medical Image Reflection for Agentic Diagnosis), a medical visual diagnostic framework designed to autonomously search for evidence and verify its relevance and consistency in image…
Parameter Exploration for RLVR via Variational Learning
Researchers propose a new approach called Perturbed Parameter Policy Optimization (3PO) to enhance exploration in large language models (LLMs). This method explores the parameter space instead of the action space…
Contextual Information Policy Optimization for Search Agents
The paper proposes a new method called Contextual Information Policy Optimization (CIPO) for search agents. This method aims to improve the efficiency of evidence use in complex or evolving information tasks by…
Contextual Information Policy Optimization for Search Agents
The text introduces a new method called Contextual Information Policy Optimization (CIPO), which aims to improve the reliability of search agents by ensuring they use retrieved evidence effectively during multi-step…
Contextual Information Policy Optimization for Search Agents
The researchers propose a new reinforcement learning framework called Contextual Information Policy Optimization (CIPO) to improve the reliability of search agents for knowledge-intensive tasks. CIPO encourages the use…
Contextual Information Policy Optimization for Search Agents
The paper proposes a new method called Contextual Information Policy Optimization (CIPO) to improve the reliability of search agents for complex or evolving information tasks. CIPO aligns policy optimization with…
EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning
Researchers have developed EnvACE, a method for training large language model agents that simulates environment interactions during training using world rehearsal instead of relying on real or synthesized executable…
EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning
The researchers introduce EnvACE, a method for reinforcement learning that replaces external environment interactions during training with world rehearsal. This method allows the policy to internalize the relationship…
EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning
Researchers have introduced EnvACE, a method for training large language model agents using world rehearsal instead of interactions with real or simulated environments. The method involves the policy alternating…
EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning
The paper introduces EnvACE, a method for training large language model agents for long-term tool use without relying on interactions with real or synthesized environments, which can be costly to construct and verify…
iARCS: Iterative Agentic RL for Controllable 3D Scene Generation
The paper presents iARCS, an iterative agentic reinforcement learning framework for enhancing the performance of synthetic 3D scene generators by adapting them to meet task-specific functional constraints, such as…
Agentic Reinforcement Learning with Observation-Calibrated Self-Distillation
Researchers propose Observation-Calibrated Self-Distillation (OCSD) to address a confounding issue in On-Policy Self-Distillation (OPSD), a method used to train large language model agents. OCSD aims to attribute…
Agentic Reinforcement Learning with Observation-Calibrated Self-Distillation
Researchers propose a new method called Observation-Calibrated Self-Distillation (OCSD) to train large language models in reinforcement learning. OCSD addresses a confounding issue in previous methods by comparing two…
TurnSight: Turn-Level Hindsight Self-Distillation for Tool-Integrated Reasoning
The authors propose a framework called TurnSight for Tool-Integrated Reasoning (TIR). This framework aims to improve fine-grained credit assignment in long-horizon TIR scenarios by addressing the limitations of…
On-Policy and Off-Policy Learning for Large Action Spaces
The paper presents research on policy learning in interactive systems with large action spaces. The study focuses on two paradigms: on-policy and off-policy learning, both of which face challenges like inefficient…
On-Policy and Off-Policy Learning for Large Action Spaces
The paper by Imad Aouali discusses policy learning in interactive systems with large action spaces, where an agent makes decisions and receives partial feedback based on a context. The work focuses on two paradigms…
On-Policy and Off-Policy Learning for Large Action Spaces
The paper discusses policy learning in interactive systems with large action spaces, where agents observe contexts, select actions, and receive partial feedback. The research focuses on two paradigms: on-policy…
OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models
The research introduces OSReward, a new benchmark for evaluating the reliability of vision-language models (VLMs) used to judge computer-using agents' trajectories. The study finds that state-of-the-art VLMs have a…
DLAM: Distributional Latent Actions with Temporal Constraints
The paper introduces DLAM, a distributional latent-action model that aims to overcome the limitations of existing models by representing each transition as a diagonal Gaussian. This model is designed to learn more…
Graph Is the Verifier: Agentic Reinforcement Learning for Interprocedural Vulnerability Detection
Researchers propose VulAgentRL, a framework for interprocedural vulnerability detection using agentic reinforcement learning (RL) and a Code Property Graph (CPG). The CPG verifies the evidence cited by the policy…
SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution
SkillRise is a new reinforcement learning framework that helps large language model agents solve related yet distinct tasks more efficiently. It organizes tasks into sequences and uses a single policy to alternate…
Compact Latent Coordination for Autonomous Vehicles at Unsignalized Intersections
A new approach called Master-Agent Proto-plan System (MAPS) is proposed to improve the coordination of autonomous vehicles at unsignalized intersections. MAPS uses a hierarchical deep reinforcement learning…
Compact Latent Coordination for Autonomous Vehicles at Unsignalized Intersections
The paper introduces a new deep reinforcement learning architecture called Master-Agent Proto-plan System (MAPS) for coordinating autonomous vehicles at unsignalized intersections. The MAPS system uses a centralized…
Compact Latent Coordination for Autonomous Vehicles at Unsignalized Intersections
Researchers present MAPS, a deep reinforcement learning architecture for coordinating autonomous vehicles at unsignalized intersections. The system uses a central Master agent that generates a global coordination…
Expert Behavior Prior Reinforcement Learning
The paper proposes a new method called Expert Behavior Prior (EBP) for behavior prior reinforcement learning (BPRL). EBP aims to improve the effectiveness of policy priors by generating expert policy priors directly…
MIRROR: Learning from the Other View for Multi-Modal Reasoning
Researchers have created a new approach called Modality-Informed Reciprocal Reasoning Optimization (MIRROR) to improve the reasoning abilities of both large language models and vision-language models. The approach uses…
PATS: Policy-Aware Training Scaffolding for Agentic Reinforcement Learning
The authors propose a new training paradigm called PATS for reinforcement learning agents that learn long-term tasks. PATS converts rollout groups from the latest policy into evidence cards and adjusts the context used…
PATS: Policy-Aware Training Scaffolding for Agentic Reinforcement Learning
The paper 'PATS: Policy-Aware Training Scaffolding for Agentic Reinforcement Learning' proposes a new training paradigm called Pats, which aims to improve the performance of long-horizon reinforcement learning agents…
PATS: Policy-Aware Training Scaffolding for Agentic Reinforcement Learning
Researchers propose a new method called PATS for training agents in long-term reinforcement learning tasks. The method uses a dynamic training scaffold that adapts to the evolving policy, providing concrete guidance to…
Active Inference as a Convex Markov Decision Process
The paper presents Active Inference (AIF) as a method that minimizes the expected free energy (EFE), combining both epistemic and pragmatic objectives within a single variational principle. The authors frame AIF as…
Co-Evolving LLM Evaluators and Policies via DynamicRubric
The researchers propose DynamicRubric, a method that allows large language models to improve by co-evolving evaluators and policies. This method generates weighted binary rubrics for each candidate set and aggregates…
Generalized Kalman filter based temporal difference reinforcement learning
The paper introduces a new reinforcement learning framework that combines conditional expectation theory with a generalized Kalman filter. This approach allows for the estimation of both the value function and its…
Towards Miniature Humanoid Tele-Loco-Manipulation Using Virtual Reality and Reinforcement Learning
The paper describes a new control stack for miniature humanoids that allows a single operator to see, manipulate, and navigate in distant environments using Virtual Reality (VR) for upper-body teleoperation and…
Measuring Reward-Seeking via Contrastive Belief Updates
Researchers have developed a method to measure 'reward-seeking' in language models trained with reinforcement learning. This behavior occurs when the model optimizes for the grader's judgment instead of the intended…
Chemical filters for ultra-high-throughput materials screening and generation
Researchers introduce a chemical validity operator to improve the reliability of AI-generated materials designs by filtering out implausible compositions that violate established chemical principles. The operator…
MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for Parameter-Efficient Reasoning in Compact Models
The paper introduces MADA-RL, a method for fine-tuning compact language models with limited budgets. It specializes the models into generator and critic roles and trains them using a debate-aware learning signal. The…
Concept-Guided Spatial Regularization for World Models in Atari Pong
Researchers evaluate five different visual world-model agents in Atari Pong and find that their rollouts contain errors such as ball disappearance and incorrect motion. They propose Concept-Guided Spatial…
Digital Pantheon: Simulating and Auditing Coalition Formation with LLM Agents
Researchers present a multi-agent framework that simulates political coalition formation using Large Language Models (LLMs). The framework combines Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO)…
Mask-Aware Policy Gradients for Diffusion Language Models
The paper presents a method for improving the performance of large language models in tasks like mathematical reasoning and coding by using reinforcement learning. This method addresses the challenge of extending…
Reachability-Aware Pretraining for Efficient Target-Oriented Path Exploration in Temporal Knowledge Graph Reasoning
The paper proposes a new method called RAPTOR for temporal knowledge graph (TKG) reasoning under the extrapolation setting. This method aims to improve the efficiency of target-oriented path exploration by using…
Reachability-Aware Pretraining for Efficient Target-Oriented Path Exploration in Temporal Knowledge Graph Reasoning
The authors propose a new method called RAPTOR for improving the efficiency of target-oriented path exploration in temporal knowledge graphs (TKG). The method injects a reachability-aware inductive bias to the agent…
Explaining Reinforcement Learning Agents via Inductive Logic Programming
This research aims to improve Reinforcement Learning (RL) policies by making them more transparent and interpretable. It proposes objective and planning-oriented metrics for policy explainability in RL settings using…
A Learning-Rate-Gated Failure of GRPO in a Small Language and Vision-Language Model Web Agent: A Controlled Null and Its Mechanism
The study investigates the impact of Reinforcement learning with verifiable rewards and Group Relative Policy Optimization (GRPO) on a small language and vision-language model web agent. The research found that GRPO…
A Learning-Rate-Gated Failure of GRPO in a Small Language and Vision-Language Model Web Agent: A Controlled Null and Its Mechanism
The research investigates the effectiveness of Reinforcement learning with verifiable rewards and Group Relative Policy Optimization (GRPO) on a small language and vision-language model web agent at scales ranging from…
Knowledge- and Gradient-Guided Reinforcement Learning for Parametrized Action Markov Decision Processes
In this research paper, the authors present a novel reinforcement learning algorithm called Knowledge- and Gradient-Guided Reinforcement Learning (KGRL). This algorithm is designed to work in Parametrized Action Markov…
OOD-RL-Bench: A Benchmark Framework for Out-of-Distribution Detection in Reinforcement Learning
The paper introduces OOD-RL-Bench, a new framework for testing out-of-distribution (OOD) detection in reinforcement learning (RL). The framework is designed to evaluate the performance of detectors against various…
TerraZero: Procedural Driving Simulation for Zero-Demonstration Self-Play at Scale
TerraZero is a procedural driving simulation developed by researchers to train autonomous driving agents efficiently at scale without the need for real-world data demonstrations. It offers improved speed, realistic…
TerraZero: Procedural Driving Simulation for Zero-Demonstration Self-Play at Scale
TerraZero is a procedural driving simulator and training stack for reinforcement learning at scale, designed to train autonomous driving agents without human demonstrations or fallback planners. The simulator provides…
Unveiling Complex Collective Behaviors from Simple Rewards
Researchers have developed a new method called the Agent Response Map (ARM) to help understand how simple rewards can lead to complex behaviors in multi-agent reinforcement learning (MARL). The ARM reveals that robots…
A Minimalist Retargeting-Guided Reinforcement Learning Recipe for Dexterous Manipulation
Researchers introduce REGRIND, a method that enables robots to learn dexterous manipulation policies from a single human demonstration. The method retargets hand-object motion to a robot reference and trains an RL…
Auditing the Risk Claims of Distributional Reinforcement Learning
In a study titled 'Auditing the Risk Claims of Distributional Reinforcement Learning', Hari Prasad investigates whether the risk claims made by distributional reinforcement learning agents are accurate. The study finds…
Learning Residual Kinematic Corrections for Continuous Neural Decoding via Reinforcement Learning
A new method using reinforcement learning (RL) has been proposed to improve the accuracy of decoding continuous three-dimensional motor imagery signals from non-invasive EEG data for brain-computer interfaces (BCIs)…
Learning Residual Kinematic Corrections for Continuous Neural Decoding via Reinforcement Learning
Researchers proposed a new method for improving the accuracy of decoding continuous three-dimensional motor imagery signals using non-invasive EEG-based BCIs. They introduced a two-stage decoding framework that applies…
ADORN: Adaptive Drift handling for Open RAN using Reinforcement Learning
The paper proposes an adaptive retraining method for Open Radio Access Networks (O-RAN) using Reinforcement Learning (RL). The goal is to balance forecasting accuracy and retraining cost in the presence of dynamic…
DrugGen 2: A disease-aware language model for enhancing drug discovery
A new generative model called DrugGen-2 has been developed to enhance drug discovery. It takes into account both disease ontology and target protein sequences, improving upon existing models in generating unique…
Multi-Modal, Multi-Environment Machine Teaching for Robust Reward Learning
The paper proposes a method for machine learning that aims to align autonomous agents' behavior with human intent in diverse operational contexts by using a hierarchical machine teaching algorithm for reward learning…
Self-Adaptive Anomaly Detection with Reinforcement Learning and Human Feedback in Connected Vehicles
The paper proposes an online anomaly detection framework for autonomous cyber-physical systems (CPS), specifically connected vehicles, that combines three coordinated mechanisms: a deep Q-network with self-attention to…
When Synthetic Speech Is All You Have: Better Call GRPO
The paper presents a new approach for adapting Large Language Model (LLM) based Automatic Speech Recognition (ASR) systems to regulated domains like banking using synthetic text-to-speech instead of real speech due to…
When Synthetic Speech Is All You Have: Better Call GRPO
Researchers present a method called Group Relative Policy Optimization (GRPO) that improves the accuracy of synthetic speech-to-text conversion compared to traditional methods. This could be particularly useful in…
Agon: Competitive Cross-Model RL with Implicit Rival Grading of Reasoning
The paper introduces Agon, a new reinforcement learning approach that trains two reasoning models by having them compete against each other. This competitive grading system aims to improve the quality of thinking…
Agon: Competitive Cross-Model RL with Implicit Rival Grading of Reasoning
The paper introduces a new method called Agon for reinforcement learning that allows two competing models to grade each other's reasoning during training. This approach aims to improve the quality of solutions by…
Reward-Adaptive Iterative Discovery: A Case Study on Automated Game Testing for NHL26
The researchers present Reward-Adaptive Iterative Discovery (RAID), a new approach for automated game testing in the gaming industry, specifically focusing on the ice hockey game EA SPORTS NHL 26. The goal is to reduce…
RLVP: Penalize the Path, Reward the Outcome
The paper 'RLVP: Penalize the Path, Reward the Outcome' by Bojie Li and Noah Shi proposes a method for training artificial agents to perform tasks effectively in real-world environments. This method addresses two main…
RLVP: Penalize the Path, Reward the Outcome
The paper 'RLVP: Penalize the Path, Reward the Outcome' discusses challenges faced by agents learning from real-world interactions and proposes a solution to improve their performance and adherence to constraints. The…
Selective Timestep Weighting and Advantage-Based Replay for Sample-Efficient Diffusion RLHF
The paper proposes two methods for improving the efficiency of Reinforcement Learning from Human Feedback (RLHF) when applied to diffusion models, which are often inefficient due to high feedback requirements. The…
Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning
The paper proposes a new method called Single-rollout Asynchronous Optimization (SAO) for Reinforcement Learning (RL) in post-training large language models (LLMs). SAO aims to improve stability and task effectiveness…
Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning
The paper introduces Single-rollout Asynchronous Optimization (SAO), a new approach to reinforcement learning (RL) aimed at post-training large language models (LLMs). SAO addresses stability and off-policy challenges…
Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning
In this paper, the authors introduce Single-rollout Asynchronous Optimization (SAO), a new approach to reinforcement learning (RL) for post-training large language models (LLMs). SAO aims to improve training stability…
Improving LLM-Generated Process Model Quality Through Reinforcement Learning: The Role of Reward Function Design
Researchers have improved the quality of process models generated by large language models (LLMs) using reinforcement learning (RL). They found that designing the reward function is crucial for optimizing beyond the…
Information Gain-based Rollout Policy Optimization: An Adaptive Tree-Structured Rollout Approach for Multi-Turn LLM Agents
The paper proposes a new method called Information Gain-based Rollout Policy Optimization (IGRPO) for improving large language model (LLM) agents in long-horizon search tasks. The method addresses the issue of existing…
Information Gain-based Rollout Policy Optimization: An Adaptive Tree-Structured Rollout Approach for Multi-Turn LLM Agents
The paper presents Information Gain-based Rollout Policy Optimization (IGRPO), a new framework for improving large language model agents in long-horizon search tasks. IGRPO aims to allocate rollout budget more…
Multi-Agent Deep Reinforcement Learning for Multi Objective Battery Management in Dairy Farms
This research introduces a multi-objective optimization control system for battery management in dairy farms that utilizes differential evolution and multi-agent deep reinforcement learning. The system aims to improve…
Adaptive Inference Batching using Policy Gradients
The paper investigates using reinforcement learning (RL) to create adaptive batching and routing policies for inference serving systems that can handle bursty, heterogeneous workloads. The research finds that in…
LLM-as-a-Verifier: A General-Purpose Verification Framework
In this research, a general-purpose verification framework called LLM-as-a-Verifier is introduced. This framework provides detailed feedback for agentic tasks without requiring additional training. The framework…
Relational Multi-Agent Reinforcement Learning for Dynamic Pricing in High-Speed Railway Markets
Researchers propose a new approach for dynamic pricing in liberalized railway systems using Relational Multi-Agent Reinforcement Learning (RL). The method models the environment as a graph, process features of entities…
TREK: Distill to Explore, Reinforce to Refine
TREK is a procedure that uses distillation for exploration support expansion to improve policy optimization on tasks where the current policy stalls on hard prompts. It identifies low-performing prompts, generates…
TREK: Distill to Explore, Reinforce to Refine
The paper proposes a method called TREK (Teacher-Routed Exploration via Forward KL) to improve Group Relative Policy Optimization (GRPO). TREK uses distillation for exploration support expansion and can work with…
Weak-to-Strong Generalization via Direct On-Policy Distillation
Researchers propose a method called Direct On-Policy Distillation (Direct-OPD) for improving stronger language models by leveraging the knowledge gained from reinforcement learning on smaller, weaker models. This…
EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments
A new evaluation setting called Autonomous Policy Evolution (APE) is introduced in a benchmark named EvoPolicyGym, which is used to test how autonomous agents iteratively improve their policy systems within a fixed…
EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments
In the research paper titled 'EvoPolicyGym', the authors introduce a controlled evaluation setting called Autonomous Policy Evolution (APE) for assessing how autonomous agents iteratively improve their executable…
WorldSample: Closed-loop Real-robot RL with World Modelling
The paper presents WorldSample, a framework that uses reinforcement learning (RL) to improve the performance of robots by creating synthetic transitions based on real rollouts and a post-trained world model. This…
Human-Machine Collaboration on Generative Meta-Learning: Model and Algorithm
Researchers have proposed a new method called Generative Meta-Learning with Human Feedback (GMHF) to improve the performance of machine learning models in environments different from their training distribution…
Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning
The paper introduces Medical Reasoning-aware Policy Optimization (MRPO), a reinforcement learning algorithm for medical multimodal reasoning, which aims to optimize the reasoning process and reduce cascading errors…
GR2 Technical Report
Researchers present GR2 (Generative Reasoning Re-Ranker), a framework designed to improve the performance of recommendation systems in an industrial setting. The system combines mid-training on semantic IDs…
QVal: Cheaply Evaluating Dense Supervision Signals for Long-Horizon LLM Agents
The paper introduces QVal, a training-free testbed for evaluating dense supervision signals in long-horizon reinforcement learning tasks. QVal measures the alignment of a method's score with the Q-values of a strong…
Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs
Researchers have developed two new mechanisms - Reinforcement Learning with Metacognitive Feedback (RLMF) and Metacognitive Data Selection - to improve the performance of Language Models (LLMs). The goal is to make…
Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs
The paper discusses a new approach called Reinforcement Learning with Metacognitive Feedback (RLMF) to improve Language Model's (LLM) metacognitive abilities. The RLMF aims to help LLMs better monitor and regulate…
TRIAGE: Role-Typed Credit Assignment for Agentic Reinforcement Learning
The paper proposes a new framework called TRIAGE for credit assignment in agentic reinforcement learning. This framework adds a semantic role axis to outcome credit, allowing for more accurate reward distribution for…
Latent Actions from Factorized Transition Effects under Agent Ambiguity
The research introduces Observed Transition Factorization (OTF), a method to decompose transitions in multi-object or distracting scenes into reusable transition primitives. These primitives are used to form…
Automating Potential-based Reward Shaping with Vision Language Model Guidance
Researchers introduce a new framework called VLM-PBRS for automating potential-based reward shaping in reinforcement learning agents. The framework learns the potential function directly from vision language model…
Automating Potential-based Reward Shaping with Vision Language Model Guidance
Researchers introduce a new method called VLM-PBRS for automating potential-based reward shaping using vision language model (VLM) feedback. The aim is to improve reinforcement learning agents' ability to learn from…
Automating Potential-based Reward Shaping with Vision Language Model Guidance
Researchers have developed a new method called VLM-PBRS for automating potential-based reward shaping in reinforcement learning agents. This method uses vision language models (VLMs) to learn the potential function…
Automating Potential-based Reward Shaping with Vision Language Model Guidance
In this research, the authors present a new framework called VLM-guided PBRS, which learns potential functions for reward shaping directly from vision language model (VLM) feedback. The goal is to address the…
Automating Potential-based Reward Shaping with Vision Language Model Guidance
Researchers have developed a new method for reinforcement learning agents using vision language models (VLMs). The approach called VLM-guided PBRS learns a potential function from VLM feedback to address the challenge…
Joint Learning of Experiential Rules and Policies for Large Language Model Agents
The paper introduces a new method called Joint Learning of Experiential Rules and Policies (JERP) for large language model agents operating in multi-step interactive environments. The main challenge addressed is…
Joint Learning of Experiential Rules and Policies for Large Language Model Agents
Researchers present a method called Joint Learning of Experiential Rules and Policies for Large Language Model Agents (JERP). This method aims to help LLM agents, which operate in multi-step interactive environments…
Learning to Fold: prizewinning solution at LeHome Challenge 2026 (1st place online, 2nd offline)
The text describes a prize-winning system for bimanual garment folding from the LeHome Challenge 2026. The system used a reinforcement learning loop to improve a vision-language-action (VLA) policy, placing first in…
Learning to Fold: prizewinning solution at LeHome Challenge 2026 (1st place online, 2nd offline)
A research paper describes a system that won first place online and second place offline in the LeHome Challenge 2026, a competition on bimanual garment folding. The system improves a vision-language-action (VLA)…
State Representation Matters in Deep Reinforcement Learning: Application to Energy Trading
Researchers studied how the representation of a state given to a reinforcement learning (RL) agent impacts energy trading decisions. They used a fixed Double DQN agent in a pumped-storage arbitrage environment and…
State Representation Matters in Deep Reinforcement Learning: Application to Energy Trading
Researchers studied the impact of state representation in deep reinforcement learning (DRL) for energy trading decisions, using a fixed Double DQN agent in the HydroDam environment. They found that combining absolute…
State Representation Matters in Deep Reinforcement Learning: Application to Energy Trading
Researchers compared different methods for representing market conditions in reinforcement learning (RL) models for energy trading. They found that combining multiple types of features led to stronger performance…
FORCE: Efficient VLA Reinforcement Fine-Tuning via Value-Calibrated Warm-up and Self-Distillation
The paper introduces FORCE, a framework designed to improve the efficiency of reinforcement learning fine-tuning for Vision-Language-Action (VLA) models. The framework consists of three stages: Value-Calibrated…
FORCE: Efficient VLA Reinforcement Fine-Tuning via Value-Calibrated Warm-up and Self-Distillation
The research introduces FORCE, a three-stage framework for fine-tuning Vision-Language-Action (VLA) models using Reinforcement Learning (RL). The framework addresses sample inefficiency issues by implementing a…
Hierarchical Reinforcement Learning for Neural Network Compression (HiReLC): Pruning and Quantization
The authors present HiReLC, a framework that uses reinforcement learning to automatically compress deep neural networks by pruning and quantizing them. The framework works in two levels: low-level agents (LLAs)…
Hierarchical Reinforcement Learning for Neural Network Compression (HiReLC): Pruning and Quantization
The paper presents HiReLC, a framework that uses reinforcement learning to compress deep neural networks by pruning and quantizing them. The framework divides the compression search into two levels: low-level agents…
Hierarchical Reinforcement Learning for Neural Network Compression (HiReLC): Pruning and Quantization
The paper introduces HiReLC, a hierarchical reinforcement learning framework that automates the joint quantization and structured pruning of deep neural networks. It operates at two levels: low-level agents (LLAs)…
Multi-Agent Goal Recognition with Team- and Goal-Conditioned Reinforcement Learning and Factorized Branch-and-Bound
Researchers propose a method called Multi-Agent Goal Recognition with Branch-and-Bound (MAGR-BB) to infer the goals of multiple agents in real applications like drone surveillance and collaborative robotics, where only…
Multi-Agent Goal Recognition with Team- and Goal-Conditioned Reinforcement Learning and Factorized Branch-and-Bound
The paper presents Multi-Agent Goal Recognition with Team- and Goal-Conditioned Reinforcement Learning and Factorized Branch-and-Bound (MAGR-BB), a method to infer which agents in a multi-agent system are working…
Semantic Consistency Policy Optimization for Reinforcement Learning of LLM Agents
The researchers propose a method called Semantic Consistency Policy Optimization (SCPO) to address a problem in reinforcement learning of LLM agents. The issue is that these agents are trained for long-term, sparse…
WinDOM: Self-Family Distillation for Small-Model GUI Grounding
Researchers have created a method called WinDOM for small GUI-grounding agents, which are useful for on-device deployment and accessibility tools. The method uses self-Family Distillation (SFD) to obtain bounding-box…
LaGO: Latent Action Guidance for Online Reinforcement Learning
The paper 'LaGO: Latent Action Guidance for Online Reinforcement Learning' proposes a new framework called LaGO. This framework uses a pre-trained language model (LLM) as a guide to help with online policy optimization…
Reinforcement Learning for Computer-Use Agents with Autonomous Evaluation
Researchers propose a framework for training Computer-Use Agents (CUAs) using reinforcement learning and autonomous vision-language evaluation. This method provides a scalable supervision signal, improving success…
Themis: An explainable AI-enabled framework for Reinforcement Learning with Human Feedback
Researchers present Themis, an AI framework that combines explainability and human feedback to train safe Reinforcement Learning (RL) systems. The framework is designed for over 200 environments, easy configuration for…
AIR: Adaptive Interleaved Reasoning with Code in MLLMs
The paper presents a new approach to enhance multimodal large language models (MLLMs) by giving them the ability to reason adaptively with code, particularly in numerical computation tasks. This is achieved through…
Decentralized Autonomous Traffic Management through Corridor Networks
Researchers propose a decentralized approach to managing traffic in Advanced Air Mobility (AAM) corridors using multi-agent reinforcement learning (MARL). The method is tested on various multi-corridor networks and…
SPIRAL: Learning to Search and Aggregate
Researchers introduced SPIRAL, a framework for improving language model reasoning at test time. The framework uses three primitives - sequential reasoning within a trace, independently sampled parallel traces, and…
VeriEvol: Scaling Multimodal Mathematical Reasoning via Verifiable Evol-Instruct
The paper introduces VeriEvol, a framework that addresses scaling issues in reinforcement learning for visual mathematical reasoning. The framework decouples the difficulty of prompts and the reliability of answers…
VeriEvol: Scaling Multimodal Mathematical Reasoning via Verifiable Evol-Instruct
The paper presents VeriEvol, a framework for scaling reinforcement learning in visual mathematical reasoning. It addresses the challenge of maintaining reliable reward labels as data volume grows by treating scaling as…
