Reinforcement Learning

Also known as RL

Follow this topic Ask about this topic

Momentum

144 papers in the last four weeks, up 243% on the four weeks before. 1.4% of all new papers.

Jul 13Week of Sep 28

Latest papers 1,138

Sep 12, 2026cs.AI

Fork Where the Model Changes Its Mind: Belief-Shift Branching for Tree-Structured Reinforcement Learning

Tree-structured rollouts give critic-free reinforcement learning with verifiable rewards (RLVR) step-level credit: fork a chain at an intermediate point, and sibling outcome differences estimate step value. Each fork adds sampling cost, so realistic budgets typically allow only a few forks per chain. A fork placed where the outcome is already largely settled yields siblings that mostly agree and provide almost no credit signal; hence, for a given tree size, where forks are placed largely determines how much step-level RL can gain. Most existing mainstream methods place forks by structure, such as fixed lengths, midpoints, and delimiters, or by next-token entropy. We formalize fork placement as locating the \emph{pivots} of the chain's value curve, where the expected outcome turns. We propose \emph{belief-shift branching}: read the model's answer belief at candidate boundaries and fork just before the step where consecutive beliefs diverge most. Three instantiations, none needing step-level supervision, span access levels: a black-box probe, a logit-lens depth profile, and a learned activation direction, which is fit offline and therefore used only in the validation before RL training. The signal only \emph{places} forks, and the probe costs about

1\%

of step compute on mathematics and under

5\%

on code when it runs inside the rollout engine. In that validation, against Monte-Carlo value curves, a belief-shift signal ranks first in each of the eight model

\times

benchmark panels, ahead of entropy, structural, and LLM-judge baselines. In RL across three model families and two domains, belief-shift forking leads every mathematics aggregate, on OLMo-3-7B by

+2.6

aggregate and

+2.9

on AIME 2026 over the strongest baseline, and sweeps every OLMo code column, by

+6.5

on LiveCodeBench-medium.

Bin Lei, Yu Li, Prafulla Kumar Choubey +7

University of Minnesota · Salesforce AI Research

Reinforcement Learning RL for Language Model Reasoning Step-Level Credit Assignment in RL Reinforcement Learning with Verifiable Rewards

Sep 11, 2026cs.LG

A Bellman Optimality Equation for Plasticity

In continual reinforcement learning, carefully managing the stability-plasticity tradeoff remains a core challenge. Recent work by Abel et al. (2025) formalized this dilemma by defining plasticity as the generalized directed information from an agent's observations to its actions, and empowerment as the generalized directed information from its actions to its observations. This formulation successfully reframes the traditional stability-plasticity tradeoff as an empowerment-plasticity tradeoff. However, while extensive literature exists on optimizing for empowerment, there is currently no research addressing the optimization of plasticity under this new definition. This paper presents preliminary work toward optimizing plasticity within Markov decision processes. We show that there exists a Bellman optimality equation for optimizing plasticity similar to previous work for empowerment.

Jeremy Lucas, Doina Precup

Mila – Quebec Artificial Intelligence Institute Montreal, Quebec, Canada

Stability-Plasticity Dilemma Reinforcement Learning Continual RL

Sep 11, 2026cs.LG

From Connectivity to Rewards: Dense Reward Learning with Directed State Graphs

The integration of graphs with Goal-Conditioned Hierarchical Reinforcement Learning (GCHRL) has received increasing attention, as graphs naturally encode task hierarchies for effective subgoal sampling. However, existing methods often overlook intrinsic connectivity information, failing to fully leverage the underlying topology for efficient learning. Most graph-based GCHRL methods use the graph as a stochastic sampling tool rather than as an environmental model that encodes connectivity and state-accessibility information. This limitation is particularly acute in quasimetric environments, where the inherent asymmetry of state transitions poses a fundamental challenge to stable policy learning and robust path planning. In this paper, we address these problems by introducing a state connectivity model designed to predict pairwise state connectivity strength in asymmetric environments. We transform these connectivity strengths into scalar auxiliary dense rewards, providing continuous guidance across multiple hierarchical levels. We demonstrate that our proposed framework, Graph-Guided Quasimetric Dense Reward (G2QDR), can theoretically be integrated into any existing GCHRL architecture, and the state connectivity model is efficiently implemented via a neural network trained on a directed state graph generated during exploration. Empirical results across a wide range of sparse reward environments indicate that, in general, G2QDR can enhance the performance of baseline GCHRL approaches with acceptable computational overhead.

Shuyuan Zhang, Zihan Wang, Xiao-Wen Chang +1

School of Computer Science, McGill University Mila – Quebec AI Institute · Google DeepMind, CIFAR Fellow

Reinforcement Learning Hierarchical RL Dense Reward RL Reward Design for RL Goal-Conditioned RL

Sep 10, 2026stat.ML

Near-Optimal Reinforcement Learning with Multi-Step Transition Lookahead

We study reinforcement learning (RL) with transition look-ahead, where the agent may observe which states would be visited upon playing any sequence of actions before deciding its course of action. Although look-ahead can substantially improve achievable performance, [1] showed that optimal planning with multi-step transition look-ahead is NP-hard. However, this hardness was established using a discount factor close to one. It was therefore unknown whether the problem remains hard for every discount factor, and whether near-optimal planning can nevertheless be performed efficiently. We resolve both questions. First, we show that for every fixed discount factor, exact planning remains NP-hard. Second, we introduce a randomized polynomial-time approximation scheme for every fixed look-ahead depth. Third, we extend our approach to account for unknown transitions. We empirically validate the soundness of our results on the wind-farm storage-control benchmark of [2], showing that our approach, optimally accounting for -step look-ahead information, offers substantially better performance than existing algorithms. [1] Corentin Pla, Hugo Richard, Marc Abeille, Nadav Merlis, Vianney Perchet : On the Hardness of Reinforcement Learning with Transition Look-Ahead [2] Chenbei Lu, Zaiwei Chen, Tongxin Li, Chenye Wu, Adam Wierman : Reinforcement Learning with Imperfect Transition Predictions: A Bellman-Jensen Approach

Corentin Pla, Hugo Richard, Marc Abeille +1

CREST, ENSAE · Criteo AI Lab FairPlay Joint Team

Reinforcement Learning Approximation Algorithms Model-Based RL

Sep 9, 2026cs.RO

Multi-Agent Reinforcement Learning for Autonomous UAV Exploration in Wildfire Response

This study develops a deep reinforcement learning framework for training Unmanned Aerial Vehicle (UAV) agents to navigate and monitor simulated wildfire environments. Results show that agents learn increasingly stable and effective behaviors over time, as demonstrated by converging loss trends, improved reward signals, and more consistent navigation patterns such as fire-boundary tracking. Overall, these findings highlight the potential of deep reinforcement learning (DRL) based UAV systems for autonomous wildfire monitoring and suggest that environmental structure and reward design influence policy effectiveness.

Caden Chandra, Jerry Ng

Aerial Robotics Reinforcement Learning Multi-Agent Reinforcement Learning

Sep 9, 2026stat.ML

Optimal Value Inference for Reinforcement Learning

We study offline inference for the optimal value in reinforcement learning under finite state and action spaces. Two new nuisances are derived as fixed points of a self-induced Bellman equation, in which we approximate the maximum Bellman operator by its softmax correspondence. We propose a debiased estimator through the Neyman orthogonality and establish its asymptotic normality under diverging horizons even when the behavior policy changes with time, as long as the nuisances have the statistical rates that can be achieved by many machine learning methods. We provide a concrete estimating procedure for these nuisances and show they can lead to valid inference. Synthetic experiments validate the numerical performance of our inference method, and we implement it in real-life decision-making problems, including bike repositioning and AI agentic tool use.

Nan Lu, Ethan Lee, James M. Robins +2

Reinforcement Learning Value Function Estimation Offline RL

Sep 9, 2026cs.GR

InstantMimic: A High Performance System for Learning Physics-based Skills in Seconds

Physics-based character control is a long-standing challenge in computer graphics and robotics, requiring policies that satisfy complex dynamics while producing realistic motion. Recent Deep RL approaches, particularly imitation learning methods such as DeepMimic, have had broad impact beyond animation, influencing robotics by enabling agile and expressive behaviors. While these approaches achieve impressive results, they remain computationally inefficient to train in practice. Despite GPU-accelerated simulation, we find that end-to-end pipelines often underutilize hardware due to overheads outside the physics solver, caused by fragmented GPU kernels and CPU memory access in the critical path. We present InstantMimic, a system that addresses these inefficiencies by making the entire training loop GPU-native. Built on a GPU-native physics backend, our unified pipeline integrates simulation, environment computation, policy inference, and policy updates within a single execution flow. As a result, InstantMimic reduces training time for diverse physics-based skills to a few seconds and makes LLM-agent-driven hyperparameter search practical.

Ikjun Choi, Geonho Leem, Jungdam Won

Seoul National University, Seoul, South Korea

Reinforcement Learning Imitation Learning Physics-Based Simulation

Sep 9, 2026cs.CL

SocialRL: Refining LLMs' Social Intelligence through Multi-turn Reinforcement Learning and Reward Design

Social intelligence enables agents to read social context, infer intent, and adapt over sustained dialogue. As language models become autonomous collaborators, it is central to building effective and trustworthy human-AI interaction. Existing reinforcement learning methods optimize single-turn utterances and sparse outcome rewards, producing short-sighted policies that struggle to manage goal-relationship tensions across multi-turn interactions. We propose SocialRL, a multi-turn reinforcement learning framework addressing both challenges. First, we apply multi-turn reinforcement learning using PPO that propagates delayed outcome rewards back to each turn, enabling long-horizon planning. Second, we design six process reward dimensions capturing the goal-relationship trade-off, including goal advancement, relational attunement, contextual coherence, etc. A reward model dynamically generates fine-grained scoring criteria for each dimension, while a stage-aware weight schedule prioritizes relationship-building in early turns, goal advancement mid-way, and balanced closure late. Across multiple social-dialogue benchmarks, SocialRL improves Goal Achievement by an average of 9.2 percentage points over the corresponding Base models. These results demonstrate the effectiveness of SocialRL across synthetic and real social scenes, as well as standard and challenging social scenarios.

Jianing Wang, Xintao Wang, Aili Chen +7

Fudan University · Hello Group

RL for Language Models Reinforcement Learning Process Reward Models

Sep 8, 2026cs.LG

PlayTrain: An Efficient Reinforcement Learning Framework for LLM-Generated Adaptable JavaScript Games

While many video-game environments (VGEs) have played crucial roles in advancing reinforcement learning (RL), developing novel VGEs or modifying existing ones to support new features, has been a laborious process requiring extensive hand-coding. Here we present PlayTrain, an RL framework that combines the abilities of large language models (LLMs) to robustly generate JavaScript (JS) games from a minimal human prompt, and an efficient pipeline that can run any JS game in a standard 'gym' environment. Not only are recent LLMs particularly good at writing JS code, but the JS format also allows users to easily play generated VGEs, while PlayTrain enables us to train RL agents on the exact same games. We demonstrate multiple use cases of PlayTrain, including cloning well-known Atari and ProcGen games in simple JS, where PlayTrain trains pixel-based agents end-to-end at over 1M agent-decisions per second on a single GPU node; and creating modified versions thereof (e.g., that support novel test sets, procedural generation logics, or game dynamics). Through PlayTrain, we reimagine RL VGE development: all we need is a single JS file, generated and modified through an LLM. We discuss promising future RL research directions that PlayTrain unlocks.

Ryan Truong, Lance Ying, Samuel J. Gershman +1

Harvard University, Cambridge, MA, USA · MIT, Department of Brain and Cognitive Sciences, Cambridge, MA, USA · Kempner Institute for the Study of Natural and Artificial Intelligence, Cambridge, MA, USA +1

Reinforcement Learning Game-Playing Agents

Sep 8, 2026cs.RO

CAST: Alternating State-Value Targets and Expanded Policy Gradients for Model-Based Reinforcement Learning

Model-based reinforcement learning (MBRL) is a family of RL methods that learn a model of the environment and use it for action selection, making it well suited to robotics due to its sample efficiency. Combining learned models with online planning can further improve action selection, as the planner can exploit the model to find better actions than the learned policy alone. Recent methods combining learned policies with online planning typically learn the value of the policy rather than the stronger planner-guided behavior. We present CAST (Critic with Alternating State-value Target), which uses planner-guided behavior to improve value learning while regularizing the value estimate with the current policy. CAST replaces the action-value critic with a state-value critic, trained using a target that combines a real planner-guided transition and an imagined transition under the current policy. The resulting value function corresponds to an alternating process between planner-guided behavior and the current policy, allowing it to benefit from the stronger planner behavior while being regularised by the policy being learned. We evaluate CAST on the DeepMind Control and HumanoidBench Suites against several state-of-the-art methods, and demonstrate successful transfer to a physical Unitree Go2 quadruped performing a dynamic handstand.

Pietro Noah Crestaz, Mohamed Yassine Kabouri, Nicolas Mansard +1

1Industrial Engineering Department, University of Trento, Trento, Italy. · 2LAAS-CNRS, Universit´e de Toulouse, CNRS, Toulouse, France. · 4Artificial and Natural Intelligence Toulouse Institute (ANITI), Toulouse. +1

RL for Robotics Reinforcement Learning Model-Based Planning Model-Based RL

Sep 8, 2026cs.RO

Learning to build covering structures with continuous adjustments

Robotic construction offers the potential to use materials more efficiently and create complex geometries, but current methods rely on rigid, high-precision plans that cannot accommodate the tolerances, inaccuracies, and unexpected changes inherent in physical fabrication. In this work, we introduce a reinforcement learning approach that forgoes predefined plans entirely, instead generating construction sequences adaptively as the structure is built. Our method operates on graph-structured state representations and a mixed (parameterized) action space, requiring both discrete block selection and continuous placement parameters. Because the stability simulation of a structure is computationally heavy, we develop an efficient exploration strategy by incorporating unilateral edges into graph neural networks, extending soft actor-critic (SAC) to this hybrid setting. We evaluate our algorithm, HSAC, against the prior method hybrid-PPO (HPPO), demonstrating significantly higher asymptotic performance and good sample efficiency. We also demonstrate HSAC's robustness to hyperparameter choices and its exploration capability, handling up to 10 discrete actions without performance degradation. Finally, we validate our approach on a physical two-robot setup, successfully building a spanning arch with 3D-printed blocks in closed-loop execution, confirming that policies trained in simulation transfer to real hardware.

Gabriel Vallat, Maryam Kamgarpour, Stefana Parascho

RL for Robotics Robotic Construction Reinforcement Learning

Sep 8, 2026cs.LG

Environments as Scaffold: Enriching Feedback to Bootstrap Self-Evolving Agents in Long-Horizon Tasks

Large Language Models demonstrate remarkable proficiency in static reasoning, yet training them as autonomous agents through Reinforcement Learning (RL) for long-horizon tasks is often hindered by severe reward sparsity. While conventional \textit{agent-side warming} up via supervised fine-tuning (SFT) can alleviate this, it is frequently limited by data scarcity and constrained exploration. To address this, we propose a paradigm shift to \textit{environment-side adaptation} by constructing \textbf{F}eedback-\textbf{E}nriched \textbf{E}nvironments (\textbf{FEEs}). Through a pilot study, we establish a feedback design strategy that reformulates environments by transitioning from action guidance to observation enrichment during the later stages of both intra-episode exploration and inter-episode evolution. Large-scale experiments on SciWorld and BFCL benchmarks using various Qwen3 model scales and RL algorithms such as GRPO, GSPO, and DAPO demonstrate that FEEs consistently yield performance improvements over standard settings. Furthermore, our analysis reveals that training with FEEs \textbf{(1)} stabilizes training dynamics by reducing entropy volatility, \textbf{(2)} facilitates proactive state-space exploration in difficult tasks, \textbf{(3) }ensures the internalization of environmental guidance into policy weights rather than acting as a mere inference-time prior, and \textbf{(4) }identifies intra-group feedback consistency as a critical boundary for stable optimization.

Hongbang Yuan, Zhuoran Jin, Yixin Cao

1Fudan University · 2CASIA · 3Shanghai Innovation Institute

Reinforcement Learning Sparse-Reward RL RL Fine-Tuning RL for LLM Agents

Sep 8, 2026cs.LG

Miles v0.1: Production-Level Post-Training

We present Miles v0.1, a full-stack, production-ready system for frontier post-training. Building upon the clean design of slime, Miles designs each stage of the reinforcement-learning (RL) training loop around a single principle: components should be verified, clean, and customizable. With accuracy, efficiency, reliability, and scalability as first-class goals, Miles aims to make frontier-scale RL accessible to researchers and enterprises alike. This report walks through the system end to end: rollout engines built on SGLang, a trainer with a choice of two backends (NVIDIA Megatron-LM and PyTorch FSDP), and three weight-synchronization transports for different deployment topologies. Beyond full-parameter RL, Miles also supports LoRA RL, on-policy distillation, supervised fine-tuning, and true-on-policy rollout-training alignment, and extends the same architecture to diffusion models. We close with an end-to-end case study: fully asynchronous agentic RL on a GLM-5.2 744B-A40B model over terminal-use coding tasks, running on 64 NVIDIA GB300 GPUs with a median step time of 263 seconds over the first 30 measured steps. Miles is open-sourced at https://github.com/radixark/miles, with the project website at https://miles.radixark.com.

RadixArk, :, Tom Chen +11

RadixArk

Reinforcement Learning LLM Agent Training RL Post-Training LLM Post-Training

Sep 8, 2026cs.AR

Routing Dense Layouts with History-Aware Offline Reinforcement Learning using LSTM

Detailed routing remains a dominant runtime bottleneck in physical design due to increasing complexity of design rules. Modern routers can struggle to resolve persistent violations under dense operating conditions. While recent work leverages reinforcement learning (RL) to dynamically select costs for each routing iteration, we find that this technique struggles with high-density designs where routing solutions are significantly harder. To address this, we present a history-aware offline RL policy which predicts iterative cost weights in these dense regimes to improve convergence across placement densities by utilizing readily available features from the router. Our policy uses conservative Q-learning similarly to prior work; however, our key insight is that addition of a lightweight LSTM architecture and additional features can retain sequence context and improve routing convergence across multiple densities and route guide qualities. Our policy can be integrated into any cost-based router with minimal pipeline changes, as it does not interfere with the core search algorithm. We evaluate our policy on held-out density and adjustment settings, including difficult operating points induced by dense placement and low guide quality. Our policy reduces design rule violations (DRVs) by an average of 92% over the top public baseline while simultaneously reducing runtime by 10%.

Afsara Khan, Austin Rovinski

New York University Brooklyn, New York, USA

Reinforcement Learning Offline RL

Sep 8, 2026cs.CL

When Metrics Reward the Worst Translations: Internalizing Cultural Reasoning for Social Media Translation Evaluation

Automatic translation quality metrics trained on general-domain corpora systematically fail on social media content, where communicative intent is encoded in culturally loaded expressions (internet slang, homophonic ciphers, and platform-specific idioms) rather than surface token patterns. We conduct a systematic empirical analysis demonstrating that standard metrics including COMET, XCOMET, and BERTScore exhibit near-zero or negative correlation with human cultural judgments, and even display a severity inversion in which scores increase as translation quality deteriorates. We further show that this failure extends to large language model judges: Qwen3-235B achieves Cohen's kappa of only 0.162, revealing that the bottleneck is not reasoning capacity but cultural grounding: models lack the domain-specific cultural knowledge needed to identify which aspects of a translation require scrutiny. To address this, we propose CuRIL, a reinforcement learning framework that internalizes cultural reasoning: cultural annotations are prepended inside the model's reasoning, excluded from policy gradients via a token-level loss mask, and injected with a probability that decays to zero over training, progressively forcing autonomous cultural judgment. On a 1,444-sample human-annotated social media translation benchmark, Qwen3-8B trained with CuRIL achieves Cohen's kappa 0.370 and Exact Match accuracy of 45.22%, approaching Gemini-3.1-Pro with 30x fewer parameters and surpassing models up to 235B in scale. We further demonstrate that our judge produces reliable reward signals for downstream translation optimization, reducing the low-quality translation rate by over 20 percentage points under independent human evaluation.

Yiwen Qiu, Linjuan Wu, Dingming Li +7

Zhejiang University · Xiaohongshu Inc.

Reinforcement Learning Cultural Knowledge in Language Models MT Evaluation

Sep 7, 2026cs.LG

CoER: Defending against Adaptive Indirect Prompt Injection via Adversarial Co-Evolution and Refinement

Language-model agents are vulnerable to indirect prompt injection (IPI) during tool use: adversarial instructions hidden in untrusted tool outputs can covertly redirect legitimate task execution. Existing work often trains and evaluates defenses against fixed attacks that do not adapt to the defender's behavior, so the resulting defenses may struggle against adaptive attacks in real-world settings. We argue that a strong defense against adaptive IPI must adapt during training to a continually evolving attacker. Building on this insight, we propose CoER, a verifier-grounded co-evolution and refinement framework that models interleaved tool calls and adaptive injections within a task as a general-sum Markov game: the defender advances the task through successive tool calls, while the attacker can inject multiple times within the same task and adapt subsequent attacks to the defender's responses and prior execution traces. After initializing the attacker from successful trajectories, bilateral adversarial reinforcement learning (BA-RL) retains historical policies from both roles as opponent populations and mixes current and historical opponents, extending training beyond the latest matchup. Attackers from these populations are then reused to challenge teacher agents, and only demonstrations verified for both safety and task completion are used to fine-tune the co-evolved defender. Across seven domains and three evaluation seeds, CoER reduces adaptive attack success from 41.3% to 0.2% and raises safe task completion from 39.6% to 76.2%; external benchmarks also show improved attack resistance. Further experiments validate the effectiveness of bilateral historical-opponent mixing and population-guided refinement. Attacker analyses show that co-evolution strengthens attack capabilities and that the trained attacker uses execution feedback to adapt subsequent injections.

Boyang Zhang, Qingxin Xiao, Lingwei Dang +1

School of Software Engineering, South China University of Technology

Reinforcement Learning AI Agent Security Indirect Prompt Injection Prompt Injection Defense Prompt Injection Attacks on LLMs Prompt Injection Attacks on AI Agents

Sep 7, 2026cs.LG

Temporal-Causal Inference for Reinforcement Learning via Automata Learning

We consider reinforcement learning in environments with dynamics that undergo an irreversible phase transition governed by a hidden temporal pattern. The agent observes the base state but cannot observe the phase directly. We formalize this problem as a two-phase non-Markovian decision process and introduce Temporal-Causal Inference for Reinforcement Learning (TCIRL), a framework that jointly learns a control policy and infers the hidden temporal cause of the phase transition. TCIRL maintains a hypothesis deterministic finite automaton (DFA) to track what phase is active and refines it via counterexample-driven SAT-based synthesis. We prove that the hypothesis converges almost surely to a DFA recognizing the true cause language on all attainable label sequences, yielding an optimal policy for the original non-Markovian decision process. Experiments on a genetic therapy gridworld and a traffic signal environment show that TCIRL recovers the correct cause DFA and matches the full-information baseline in both domains.

Jan Corazza, Daniil Kaminskyi, Simon Lutz +4

Research Center Trustworthy Data Science and Security, TU Dortmund University, Dortmund, 44227, Germany. · School for Engineering of Matter, Transport, and Energy at Arizona State University, Tempe, Az 85281, USA.

Reinforcement Learning Causal RL

Sep 7, 2026cs.LG

Long-Horizon Language Model Reinforcement Learning via Progressive Point Matching

Current paradigms for training language models via reinforcement learning rely heavily on sparse outcome rewards. However, as we pursue tasks that require longer and more complicated trajectories, such strategies result in slow learning. Prior work has attempted to address this problem by rewarding partial progress; however, naive formulations are often biased and converge to suboptimal policies. We show that a simple and unbiased dense reward formulation, which we term progressive point matching, scales exponentially more efficiently to long-horizon tasks by rewarding partial progress on a segment level, both theoretically and empirically via synthetic environments. We then show how progressive point matching can be practically instantiated using a single reference trajectory per task. On extremely hard math reasoning problems, sparse outcome rewards cannot make any progress, whereas segment-level rewards enable improvements at larger test-time token budgets when measured by success rate or pass@k.

Preston Fu, Kevin Frans, Oleh Rybkin +2

UC Berkeley · Carnegie Mellon University

Reinforcement Learning Reward Shaping RL for Language Model Reasoning Dense Reward RL LLM Reasoning

Sep 7, 2026cs.CL

CircuitLens: Reasoning Circuits as Data Selection Signals for Reinforcement Learning with Verifiable Rewards

Reinforcement learning with verifiable rewards (RLVR) is sensitive to which problems a model trains on, yet existing selection criteria--difficulty filtering, hand-curation, reward-trajectory scoring--assess data value as an intrinsic property of problems, independent of the model that will learn from them. We introduce Circuit Reasoning Score (CRS), a selection signal derived from 46 reasoning-sensitive attention heads identified via contrastive ablation, computed in a single forward pass on the frozen base model without reward labels or rollouts. CRS runs against the intuitive hypothesis that stronger reasoning-circuit engagement produces better training data: on Qwen2.5-Math-7B, the lowest-engagement decile improves over random selection on three medium-difficulty benchmarks (GSM8K +2.0 pp, OlympiadBench +1.6 pp, Minerva +2.9 pp), while the highest-engagement decile gains less and is indistinguishable from the middle decile. The advantage has boundary conditions: on a domain-curated pool no selection method separates from the others; at 1.5B scale the useful direction differs; and the lowest-reward training condition produces the strongest downstream generalization. Within the Qwen2.5-Math settings tested, RLVR data selection appears regime-dependent rather than reducible to a static ranking of problem quality.

Zhuofan Chen, Ziqian Jiao, Yikai Cui +3

School of Computer Science and Engineering, Beihang University, Beijing, China · Beijing Institute for General Artificial Intelligence, China

Numerical Reasoning in Language Models Reinforcement Learning Reinforcement Learning with Verifiable Rewards

Sep 7, 2026cs.LG

Stable-MM-R1: Anchoring Multimodal Reasoning Dynamics via Entropy-Guided Stratification

While Reinforcement Learning (RL) effectively incentivizes reasoning in Large Language Models, current pipelines are hindered by training instability and rapid entropy collapse. These limitations often stem from "Rollout Silencing" and low-quality gradient signals in standard sampling procedures. In this work, we propose a robust, data-centric framework to stabilize RL training. We first introduce Potential-Aware Query Mining (PAQM), which filters data dynamically to focus on the "Distillation Zone"---samples with high potential for capability elicitation. Furthermore, we present Hybrid Stratified Replay (HSR), a novel mechanism that restructures batches by stratifying rollouts based on Path Entropy, a rollout-level confidence proxy, and outcome reward. Within each optimization step, HSR reuses current-policy "Stability Anchors" and "Hard Negatives" to construct high-contrast optimization groups, then clears its buffers before the next step. This approach mitigates entropy collapse while improving the utilization of learning signals under limited compute. Our method outperforms strong baselines on complex reasoning tasks, offering a principled solution for stable and efficient RL fine-tuning.

Yimeng Ye, Shuang Chen, Wenxuan Huang +8

Columbia University · CUHK · THU +1

Reinforcement Learning RL Fine-Tuning Sample-Efficient RL RL for VLMs

Sep 7, 2026cs.AI

SQL-Zero: Self-Evolving Text-to-SQL

Training a competitive Text-to-SQL agent usually depends on human-annotated natural-language/SQL pairs, which are expensive, domain-specific, and a bottleneck for scaling to new databases. We show it is possible to train a competitive solver with zero annotated pairs. We introduce SQL-Zero, a proposer-solver self-play in which a challenger and a solver start from the same base LLM and the only ground truth is execution against the database itself. The challenger generates SQL pairs calibrated to the solver's current difficulty (targeting "hard but solvable"), and both roles are updated with GRPO in alternating turns, with a template-level repetition penalty on the challenger to prevent diversity collapse. Training on BIRD databases with no labels, self-play improves over the zero-shot base on BIRD dev by 6.6 points at 3B and 7.3 points at 7B. It also scores higher than a matched control trained under the same recipe on human BIRD gold over the same databases, although an exact paired test does not resolve that margin. Transfer depends on scale: at 3B every iteration outperforms the base on unseen Spider databases and under lexical perturbation (Spider-Syn), where it also degrades less than the matched BIRD-gold control, whereas at 7B only the first iteration preserves transfer.

Daniel Machado Pedrozo, Julia Soares Dollis, Bryan Lincoln Marques de Oliveira +3

Instituto de Informática, Universidade Federal de Goiás Goiânia, GO, Brazil

Reinforcement Learning Self-Play RL LLM Agent Self-Improvement Text-to-SQL

Sep 3, 2026cs.CR

SENTINEL-RL: Offloading Topological Reasoning from LLM Agents in the Security Operations Center

Large language model (LLM) agents are increasingly proposed as autonomous SOC analysts, but two limitations make them unreliable at enterprise scale: a finite context window cannot hold a multi-thousand-host authentication graph, and free-form generation offers no guarantee that a recommended containment action is consistent with the topology it operates on. We present Sentinel-RL, an agentic-SOC architecture that decouples topological reasoning from semantic reasoning: a heterogeneous graph attention encoder summarizes the live authentication subgraph into a fixed-dimensional state, a Proximal Policy Optimization (PPO) policy maps this state to a constrained set of investigative actions, and an LLM agent loop is restricted to consuming the policy's recommendations and producing analyst-readable narratives gated by a critic. We instantiate the system on the LANL Comprehensive, Multi-Source Cyber-Security Events dataset and the Indiana University Quartz HPC cluster, reporting four results: (i) a two-phase CREATE ingestion pattern loads a 24M-edge authentication subgraph into Neo4j in 14.2 minutes on a single 32-core node, roughly 24x faster than the canonical MERGE-based pipeline; (ii) a sliding-window alert engine reliably trips a 25-event/10-second threshold in <=2.5 s across 50 trials; (iii) PPO training over 200 iterations converges to a mean episodic return of 8.74+/-0.31, with held-out precision of 0.91 and recall of 0.87 on labeled red-team events; and (iv) the integrated containment loop completes a full detect-investigate-recommend-human-approve cycle in a median of 6.3 s. We contribute a reusable engineering pattern (the hot-node deadlock workaround), a portable HPC deployment pattern (anchor-node co-location), and an enterprise-readiness analysis covering false-positive economics, reversibility guarantees, audit compliance, and the human-approval boundary.

Uday Vallabhaneni, Cassie L. Cagwin, David J. Wild

Luddy School of Informatics, Computing and Engineering · Indiana University, Bloomington, IN, USA

Reinforcement Learning Cybersecurity LLM Agent Security

Sep 3, 2026cs.AI

Spurious Advantage Hidden in GRPO

Group Relative Policy Optimization (GRPO) is widely studied for reinforcement learning with verifiable rewards, where its advantage estimator assigns each rollout a magnitude from within-group reward statistics. In the common case, this magnitude rewards rollouts that reach the correct answer through reasoning. Yet, an overlooked case shares the same surface: a rollout may land on it by guessing, and the formula still assigns a high magnitude, which we identify as the spurious advantage. This arises in three cases: bounded-answer tasks with a small candidate set; open-answer sets hosting bounded sub-cases; and search agents whose budget opens many paths to the same answer. In all three, this misleads the policy toward guess-like behaviors. We propose SIGNBALANCE, whose magnitude is composition-free: it keeps the verifier sign, uses a global scale, and restores zero-mean balance via a stop-gradient per-class rescaling. Across math and search agent benchmarks at different scales, SIGNBALANCE matches GRPO on open-answer math and improves on bounded-answer math and search agents. Code will be released.

Jiamian Wang, Samyadeep Basu, Koustava Goswami +2

Rochester Institute of Technology · Adobe Research

Reinforcement Learning Group Relative Policy Optimization Reinforcement Learning with Verifiable Rewards Policy Learning Reward Design for RL

Sep 3, 2026cs.CL

CROCODIL: Cross-Model Code Editing with LLMs

Large language models (LLMs) have become ubiquitous tools for code generation and editing. However, development teams often use multiple LLM assistants. Different developers may prefer different models, and individual developers may switch between models across different coding sessions. Because of this, the edits any one model makes are frequently applied to foreign code originally generated by another model. These LLMs are often trained on different datasets, and as a result have different stylistic preferences. Do LLMs behave differently when they edit foreign code originally written by a different LLM with a different coding style? We find that models tend to make more, and often excessive, edits on foreign code. We introduce CROCODIL (Cross-model Code Editing with LLMs), a post-training framework for reducing excessive edits while preserving functional correctness. CROCODIL's similarity reward penalizes large changes, while its execution reward scores build and test success. We use the product of these two rewards to encourage the policy to decrease the edit size without decreasing the edit task success rate. CROCODIL is available at https://github.com/EngineeringSoftware/Crocodil.

Linghan Zhong, Aditya Thimmaiah, Jayanth Srinivasa +2

1The University of Texas at Austin · 2Cisco Research

Reinforcement Learning LLM Post-Training Code Language Models

Sep 3, 2026cs.LG

Out-of-Distribution Generalisation with Sequence Models in Offline Multi-Agent Reinforcement Learning

Generalising to unseen tasks remains a fundamental challenge in offline multi-agent reinforcement learning (MARL). In this work, we present a principled analysis of zero-shot task generalisation in the offline setting and conduct an extensive empirical investigation into the scaling behaviour governing task diversity, dataset size, and network capacity. To facilitate this study, we extend offline sequence modelling architectures to handle multi-task observation and action spaces alongside variable agent counts across tasks. Our primary finding is that scaling task diversity---rather than sheer dataset size is the dominant factor in achieving robust zero-shot transfer. Through large-scale experiments across four challenging environments (Connector, RWARE, SMAX, and LBF), we demonstrate that our multi-task approach achieves a mean improvement of 3.2x on held-out test tasks compared to single-task models and consistently outperforms strong behaviour cloning baselines. These results suggest that the development of generalisable MARL agents should prioritise the diversity of the training distribution with varying numbers of agents, providing a roadmap for scaling offline MARL effectively.

Oussama Hidaoui, Omer Ebead, Ulrich Armel Mbou Sob +14

InstaDeep · AIMS · Stellenbosch University

Zero-Shot Learning Reinforcement Learning OOD Generalization Multi-Task RL Offline MARL

Sep 2, 2026cs.LG

A Comparative Study of Graph Representations for GNN-Based Power Grid Control in L2RPN

Graph construction is a critical but underexamined design choice in deep reinforcement learning for power grid control. We present a controlled experimental comparison of different graph representations, including physical topology, electrical-sensitivity, and hybrid variants for topology control in the Learning to Run a Power Network (L2RPN) environment. Our findings indicate that matching graph complexity to task granularity is more important than maximizing representational richness, and highlight the importance of controlled representation studies at scale.

Adrian Degenkolb, Qiong Huang, Benjamin Schäfer

Karlsruhe Institute of Technology Karlsruhe, Germany

Power Systems Graph Neural Networks Reinforcement Learning

Sep 2, 2026cs.CL

NE-R1: Enhancing Named Entity Recognition Model via Reinforcement Learning

Named Entity Recognition (NER) has achieved substantial progress since the advent of large language models (LLMs). Nevertheless, the recognition of long-tail and domain-specific entities remains challenging due to the deficiency in parametric knowledge. Retrieval-augmented generation (RAG) offers a promising remedy by injecting external knowledge, but it also introduces noise and unnecessary cost when dealing with familiar cases. In this paper, we propose NE-R1, a novel framework for adaptive retrieval-augmented NER. We design a "retrieval-on-demand" mechanism for NER. Then we integrate it into models by a two-stage training method: (1) multi-task instruction tuning initialization; (2) end-to-end RL optimization with CoT. To achieve reasonable selection between parameterized and external knowledge, we design a multi-dimensional reward considering both accuracy and retrieval benefit. NE-R1 achieves state-of-the-art performance on various benchmarks, with an average F1 score gain of 2.52% in in-domain evaluation and 1.18% in zero-shot cross-domain evaluation.

Meixuan Chen, Hehan Li, Ruizhi Zhao +8

Shenzhen Graduate School, Peking University · Baidu, Inc., Beijing, China · Shenzhen University, Shenzhen, China +3

Named Entity Recognition Adaptive RAG Reinforcement Learning

Sep 2, 2026cs.LG

Recursive Value Learning for Long-Horizon Offline Goal-Conditioned RL

Scaling offline goal-conditioned reinforcement learning (GCRL) to long-horizon tasks is difficult because (1) long-range value learning depends on shorter-range estimates that may still be inaccurate, and (2) max-based value backups can amplify overestimation through repeated propagation. We propose DCRL (Divide-and-Conquer RL), which recursively decomposes each trajectory segment into a balanced binary tree and trains the values from leaves to root. Each parent is therefore updated only after its children, using an exact factorization of the observed route rather than selecting among noisy alternatives. Since this objective learns values along demonstrated routes that are not necessarily optimal, DCRL jointly propagates values across trajectories to discover shorter routes. Thanks to the balanced binary tree, DCRL reduces worst-case bootstrap depth from linear to logarithmic, and this shorter dependency structure empirically corresponds to much slower error accumulation. Across diverse goal-reaching tasks, DCRL substantially outperforms prior flat offline GCRL methods, and on the five most challenging long-horizon OGBench tasks, it improves the best prior average score from 55 to 64, surpassing all flat and hierarchical baselines.

Hyeonseong Jeon, Youngwoon Lee

1Yonsei University · 2Seoul National University

Reinforcement Learning Offline RL Value Function Estimation Hierarchical RL

Sep 1, 2026cs.LG

On-Policy Distillation Meets Off-Policy GRPO: Training Compact Instruction-Following Rerankers

Compact instruction-following rerankers are attractive for deployment, but conventional distillation pipelines typically train students by offline imitation of teacher outputs on a fixed set of examples, constraining supervision to the teacher's observed ranking space. We revisit reranker distillation through the lens of reinforcement learning. We propose a two-stage framework combining off-policy teacher optimization with on-policy student distillation. In Stage 1, a 4B teacher reranker is strengthened with off-policy GRPO using LLM-judge feedback on 88K instruction-following examples. In Stage 2, a compact 1B student samples rankings from its own policy and receives soft teacher-derived rewards on those rankings, coupling student exploration with knowledge transfer. Our strongest gains appear under distribution shift. On MAIR-11, the original 11-subset, 869-query evaluation, the proposed student reaches 0.7670 nDCG@6, outperforming offline listwise KD by +4.6 points. Controlled comparisons against offline pairwise RankNet KD and on-policy GKD show that neither changing the offline distillation objective nor moving teacher-distribution matching on-policy reproduces the performance of reward-based on-policy distillation over student-sampled rankings. The advantage persists on MAIR-Full: across all 126 tasks and 9,356 queries, the proposed method obtains the highest task-macro point estimates among the evaluated distillation variants, reaching 0.6808 nDCG@6 and 0.7865 MRR@6. It also exceeds two released 7B RL-trained rerankers on the comparable MAIR-11 evaluation, while the same Stage 2 training procedure consistently improves three architecturally distinct alternative student backbones. On the 9,861-query validation benchmark, the resulting 1B reranker achieves 0.7624 nDCG@6 while providing a favorable quality-efficiency tradeoff relative to larger alternatives.

Vignesh Prabhakar, Jialing Pan, Anil Babu Ankisettipalli

SAP Labs, Palo Alto, CA, USA

Reinforcement Learning Learning to Rank LLM Reranking On-Policy Distillation

Sep 1, 2026cs.CL

The Rise of Verbal Reinforcement Learning

Natural language is emerging as a primary feedback channel for improving language agents, capable of conveying intent, preferences, and causal structure in forms interpretable by both humans and modern language models. We call this paradigm Verbal Reinforcement Learning (VRL) and offer the first unified account of it. We organize the field around a single axis, \textit{when} verbal feedback takes effect in an agent's lifecycle and \textit{what} it modifies, yielding three pillars: (1) \textbf{Language as Grounding Signal}, where language defines the task itself by specifying goals, states, and reward structures; (2) \textbf{Language as Deliberative Feedback}, where natural language guides reasoning at test time without the need to update model parameters; (3) \textbf{Language as Learning Signal}, where language-based feedback shapes model parameters through training. Within each pillar, we synthesize representative work, distinguish key subcategories of approaches, and outline the distinct role language plays in shaping agent behavior. Together, this taxonomy shows how verbal reinforcement is reshaping agent development, while also defining the challenges and opportunities for building more capable and aligned agents.

Kshitij Tayal, Arun Sharma, Genta Indra Winata +2

1AI Foundations, Capital One · University of Minnesota, USA

RL for Language Models Reinforcement Learning RL for LLM Agents