Policy Learning

Follow this topic Ask about this topic

Momentum

79 papers in the last four weeks, up 316% on the four weeks before. 0.8% of all new papers.

Jul 13Week of Sep 28

Latest papers 416

Sep 23, 2026cs.RO

Uncertainty-Gated Exploration Noise Suppresses Task Collapse in Online RL Fine-Tuning of a Flow-Matching Vision-Language-Action Policy

Online reinforcement learning fine-tuning of pretrained flow-matching vision-language-action (VLA) policies promises robots that keep learning after deployment, but continued updates often destroy competence on individual tasks while the aggregate still looks healthy. We study this failure mode, which we call task collapse, under a matched small-compute budget on LIBERO-10 with a 450M-parameter SmolVLA policy trained by PPO with stochastic (SDE) sampling. Three exploration-noise policies differ in one live variable: a fixed noise scale, a ReinFlow-style learned noise network, and an uncertainty-gated controller that redistributes exploration across task streams from task-agnostic novelty and competence signals, without task labels or episode boundaries. Under the pooled definition, fixed noise collapses tasks in two of three seeds and learned noise in every seed measured to iteration 200, while the controller collapses none in any of its three seeds. Measured parameter displacement shows the controller's action expert keeps changing, while its mean applied noise is close to the fixed scale in the available logs. The matched comparison supports the controller's effect on task preservation; the separate contributions of its adaptation across states and over time are not disentangled. A lower fixed scale slows the decline but does not stop it. No arm improves on the behavior-cloning baseline in this budget. Two properties of that regime are measured beside this result, not offered as its cause: following the reference recipe, training runs in bfloat16 with no fp32 master copy, under which 96.02% of the action expert's elements stay bit-identical across three consecutive iterations, and an fp32 master copy at the reference learning rate collapses both arms in a single-seed observation. We release tools measuring per-task collapse under four definitions, rescoring noise and instrument tares.

Mehmet Turan Yardımcı, Yunus Emre Çoğurcu

Karlsruhe Institute of Technology · Çukurova University

Robot Policy Learning RL Exploration Vision-Language-Action Models Multi-Task RL RL Fine-Tuning Policy Learning

Sep 22, 2026cs.AI

Reinforcement Learning with Decomposed Subtasks

Group Relative Policy Optimization (GRPO) and related policy-gradient methods for training language model agents collapse an entire multi-turn rollout into a single scalar trajectory reward before it enters the policy update. When the task composes distinct skills, especially under sparse and delayed environmental feedback, this collapsing is lossy: the optimizer must implicitly infer which competency drove the outcome and how that should change behavior. We argue the right primitive is not a better scalar but a decomposition: trajectory reward should be split along subtasks before it enters the policy update. We introduce Reinforcement Learning with Decomposed Subtasks (RLDS), whose core is Subtask-Decomposed Advantage Estimation (SDAE): a replacement for the scalar GRPO advantage that splits trajectory reward into per-subtask shares on a fixed taxonomy, computes a group-relative advantage per subtask, and distributes per-token credit by weighting each subtask's advantage by its importance, concentrating it around the step where a reflection marks that subtask's execution as consequential. We evaluate on four agentic benchmarks: FrozenLake (sparse grid navigation), HotpotQA (multi-hop QA, one retrieval tool), ScienceWorld (long-horizon embodied science), and DeepResearch (long-form research, four tools, composite rubric reward). Heterogeneity diagnostics emitted during training show where decomposition pays off - gains scale with subtask heterogeneity, largest on the high-heterogeneity tasks ScienceWorld (+11.5 points, paired-bootstrap 95% CI [+9.8, +13.3]) and FrozenLake (+9.8 points, [+7.0, +12.8]), and within noise on HotpotQA and DeepResearch, where the diagnostics predicted little to recover. ScienceWorld is also more compute-efficient under RLDS than scalar GRPO (-10.9% wall-clock per step), as long rollouts amortize the fixed reflect-and-grade overhead.

Mattie Terzolo, Mikolaj Sacha, Ayan Sinha +1

Upwork

Group Relative Policy Optimization Credit Assignment in RL Policy Learning Reward Design for RL RL for LLM Agents

Sep 22, 2026cs.LG

PACT: From Credit Assignment to Critic Alignment

Reinforcement learning has become a central component of large language model (LLM) post-training, yet token-level credit lacks a generally accepted mathematical definition, leaving its relationship to commonly used training signals unclear. We formulate three regularity conditions, namely Completeness, Prefix Consistency, and Neutrality, and prove that they uniquely determine token-level credit. This characterization provides a unified basis for explaining phenomena across existing algorithms and guides the development of an improved actor-critic training procedure. Through this lens, an ideal teacher in On-Policy Distillation (OPD) acts as an implicit critic, yielding an expected policy gradient proportional to that induced by token-level credit. Response-level REINFORCE Leave-One-Out (RLOO) signals match the expected policy-gradient contribution of token-level credit despite their coarser granularity. We further establish approximate credit sparsity under bounded outcome rewards and show how intermediate critic errors in Generalized Advantage Estimation (GAE) can become comparable to the underlying credit. These motivate Policy Aligned Critic Training (PACT), which adopts an Actor-then-Critic update order to apply importance sampling correction to critic training and better align the critic with the updated policy. In agentic mathematical reasoning, PACT achieves 72.87% average accuracy across four benchmarks, outperforming GRPO and PPO by 8.80 and 13.16 percentage points, respectively. On SWE-bench Verified, PACT achieves a pass rate of 67.4%, outperforming PPO, GRPO, and SAO by 2.4, 2.0, and 3.8 percentage points, respectively.

Jiayan Fu, Hang Xu, Yong Zhang +4

AllSpark Team

Reinforcement Learning Token-Level Credit Assignment RL for Language Model Reasoning Step-Level Credit Assignment in RL RL Post-Training Actor-Critic Methods+1

Sep 21, 2026cs.RO

MimicAgent: Quadruped Skills via Prompt-to-Trajectory Generation

We present MimicAgent, a prompt-to-trajectory generation framework for learning dynamic quadruped skills. Although reward shaping is extensively used when training quadruped policies, navigating the resulting reward landscape is notoriously difficult, requiring hours of "graduate student descent". Eureka attempts to automate reward design with LLMs, but we find that it struggles to generalize across diverse skills and morphologies. Our key observation is that it is far easier for a human - and by association, an LLM - to generate reference motions than to shape reward functions. Our hypothesis is motivated by the success of example-guided RL for humanoids, which exploits large-scale motion capture datasets as references for training locomotion policies. Unlike humanoids, quadrupeds lack such reference motion data. Towards this end, we propose MimicAgent, an agentic harness that, given a skill prompt, generates quadruped reference trajectories with coding agents. These coarse reference trajectories are then used to train example-guided RL policies that are deployable in simulation and in the real-world. Notably, we find that when prompting Claude Fable 5.1 within our agentic harness, 87% of prompts yield semantically aligned reference trajectories.

Lucky Kant Nayak, Narayanan Palghat Parameswaran, Neehar Peri +1

Carnegie Mellon University

Reinforcement Learning Trajectory Generation LLM Agent Trajectory Synthesis RL for Legged Locomotion Policy Learning Quadruped Locomotion

Sep 18, 2026cs.RO

SeeQ: Training Generalist Value Functions for Long-Horizon Robotic Manipulation

Despite rapid progress, generalist robot policies remain brittle on complex, long-horizon tasks that comprise multiple stages or require repeated attempts and deliberation on the same underlying stage before success. Q-value functions can improve these policies by ranking candidate actions or guiding policy improvement, but learning from sparse task-level rewards entails long credit-assignment horizons, difficult Bellman backups, and broad data-coverage requirements. We introduce SeeQ (Subtask-elicited Q-functions), which instead learns Q-values for the currently active subtask. This shortens the value-prediction horizon and enables effective learning with temporal-difference (TD) objectives. During training, subtask-level annotations present in offline robot data provide the decomposition and enable learning from broad, potentially suboptimal robot datasets. To eliminate the need for human annotations or modular subtask prediction systems at test time, our Q-function architecture is trained to autoregressively predict the active subtask in natural language before estimating its value. We instantiate SeeQ using a base vision-language backbone, pretrain it on diverse open-source robot manipulation data, and finetune it on downstream tasks. Across four real-world manipulation tasks on two bimanual robot platforms, the SeeQ value function substantially improves best-of-N policy steering.

Saksham Singh, Zheyuan Hu, Max Sobol Mark +3

Carnegie Mellon University

Robot Policy Learning Long-Horizon Robotic Manipulation Robotic RL VLMs for Robotics Policy Learning

Sep 17, 2026cs.RO

Workspace Models: Lightweight Robotic Memory via Saliency-Driven Supervision

Complex robotic manipulation tasks frequently require a long-term memory of past events and actions. As conditioning on full histories renders policies prone to spurious correlations and degrades performance, many approaches to policy memory involve compressing historical information through expensive VLM queries in-the-loop to process only task-salient information. In this paper, we propose an alternative approach in which computationally intensive VLM queries are made during train-time to learn a lightweight latent memory that can be efficiently queried at deployment time. Our representation, which we call the workspace token, is trained by (1) using a VLM to identify current and historical information necessary for completing a task, then (2) distilling these into the workspace token using a set-reconstruction decoder loss. In both simulation and hardware, we show that the workspace token can be used as a drop-in replacement for observations during deployment, enabling policies to solve memory-intensive tasks without the need for VLM reasoning in-the-loop, in effect serving as a latent harness for distilling a stronger reasoning models ability to solve long-horizon tasks to a reactive robotic policy. We further demonstrate that the workspace tokens are not only more lightweight, but also lead to better policy performance compared to conditioning policies on explicit modalities like curated past image frames, motivating a latent approach to history curation and reasoning model harnesses more broadly.

Nitish Dashora, Douglas Chen, Idan Shenfeld +3

Massachusetts Institute of Technology · Carnegie Mellon University

Robot Policy Learning Long-Horizon Robotic Manipulation Policy Learning Memory-Augmented Agents

Sep 17, 2026cs.RO

Accelerating Visual Policy Learning with Sampling-Based Model Predictive Control

Learning visual policies for locomotion and manipulation requires coordinating contact with the environment and can incur substantial computation and GPU memory costs. First-order policy gradients (FoPG) reduce training cost through differentiable simulation, but local optimization can converge to unintended contact patterns. To address this shortfall, we propose Sampling-Guided Policy Search (SGPS), which couples recurring action-target refinement by sampling-based model-predictive control with first-order policy optimization. Behavior cloning initializes the policy from sampled actions; training then alternates sampling-based refinement with short-horizon FoPG updates under perturbed initial states and randomized dynamics. For visual policy training, we use a decoupled FoPG formulation that excludes rendering from the computation graph, enabling direct learning from depth observations without a state-policy teacher. On a single GPU, SGPS learns policies for locomotion, obstacle traversal, crate pushing, and bimanual carrying on simulated Unitree Go2 and G1 robots. Our experiments further show that refinement improves policy learning beyond initialization and tracking alone. For hardware deployment, the distilled policy transfers zero-shot to a real Go2 and uses onboard depth to autonomously trot, crawl, clear hurdles, and switch between these behaviors.

Yilang Liu, Haoxiang You, Qian Wang +2

Department of Mechanical Engineering, Yale University, New Haven, CT 06520, USA · Department of Computer Science, Yale University, New Haven, CT 06520, USA · School of Electrical and Computer Engineering, University of Sydney, Sydney, NSW 2006, Australia

Policy Optimization Visuomotor Policy Learning Model Predictive Control Robotic Manipulation Robot Locomotion Policy Learning

Sep 17, 2026cs.LG

MATCH: Model-Aware Tool Learning with Curriculum Scheduling and Hierarchically Gated Rewards

Tool learning enables large language models (LLMs) to use external tools for tasks beyond parametric knowledge. Reinforcement learning can optimize tool-call behavior from feedback, but current methods still face two problems: fixed-threshold curricula can become misaligned with the policy's evolving capability boundary, and additive rewards can leak argument-level credit when the predicted tool is wrong. To address these problems, we propose MATCH, a closed-loop framework for model-aware tool learning with curriculum scheduling and hierarchically gated rewards. Model-Aware Curriculum Learning (MACL) maintains reward-derived sample difficulty that co-evolves with the policy, and each epoch selects samples near the current capability boundary together with a top-k pool of harder cases. Hierarchical Tool-call Gated Reward (HTGR) scores tool name, argument key, and argument value as a gated chain, granting credit at each level only when prerequisites hold. The same HTGR rewards drive both GRPO updates and MACL's difficulty refresh, closing the loop between policy optimization and sample scheduling. On API-Bank and BFCL V3, MATCH reaches 72.19% and 62.87% overall accuracy, outperforming the main supervised and RL-based baselines. Backbone experiments further show consistent improvements across four backbones from two model families.

Shihao Liu, Hao Yin, Lijun Liu +3

School of Cyber Security, University of the Chinese Academy of Sciences · Honor Device Co., Ltd · University of the Chinese Academy of Sciences

Reward Modeling Curriculum RL Policy Learning Tool Use RL for Tool Use

Sep 17, 2026cs.LG

EPIG-Tree: Compute-Optimal Branching for Gradient-Efficient Reinforcement Learning

Reward-based reinforcement learning for language models, exemplified by Group Relative Policy Optimization (GRPO), collapses an entire stochastic trajectory into a single scalar reward. This is clean and scalable, but it explores and allocates reward inefficiently: a trajectory may contain many causal decisions, recovery attempts, and environment-randomness events, yet every token or action inherits one trajectory-level advantage. We study tree-based rollout construction as a compute-allocation problem for policy-gradient estimation. Our central claim is that branches should be placed not where the policy is merely uncertain, but where an additional branch most reduces uncertainty about the policy gradient per unit of compute. From a law-of-total-variance decomposition of the local policy-gradient random variable, we derive two allocation laws: new branches reduce decision uncertainty, while repeated suffix rollouts reduce continuation uncertainty. The resulting EPIG-Tree score allocates branches using the already computed rollouts. It estimates occupancy- and score-weighted value uncertainty, along with a suffix law

n_e \propto w_e \|\nabla_θ\log π(a_e|h_e)\| σ_e / \sqrt{c_e}

. Empirically, EPIG reduces gradient MSE in cloned-state control, winning in all nine dense continuous-control environments of a 13-environment sweep and recovering the reference gradient direction near-perfectly, and it improves frozen-LLM gradient calibration relative to entropy branching. In online single-turn math, tree-local credit beats flat GRPO, while branch placement is secondary to token-level credit assignment. In online multi-turn Wordle, EPIG attains the highest final win rate (0.850), overtaking flat GRPO, which saturates early at 0.790, and entropy branching as training proceeds, confirming that the gradient-estimation advantage transfers to a stateful, large-action setting.

Nikita Khomich, Leopold Hermansson, Ido Hakimi

RL for Language Models Policy Gradient Credit Assignment in RL Policy Learning

Sep 17, 2026cs.RO

Learning and Transferring Closed-Loop Robot Software

Closed-loop robot policies require observation processing, state management, and situation-dependent branching, making them costly to design and tune manually. Although coding agents increasingly support control-code generation and optimization, it remains unclear whether implementations improved on source tasks also support policy acquisition for new tasks. We study this question by treating complete closed-loop implementations as reusable execution experience. For each source task, a coding agent generates policy code from a few successful demonstrations and iteratively improves it using simulation feedback. The validation-selected implementations are retained in a software archive. For new tasks, the agent generates and improves policies using archived implementations, target demonstrations, and execution feedback. The resulting policy is then frozen and executes without further model calls. Across four source tasks in RoboCasa, iterative optimization increases mean success from 28.3% to 64.2%. Across nine target tasks and three independent runs, mean success is 45.2% without references, 41.5% with initial source code, and 57.0% with optimized source code. Optimized references outperform initial references in all three runs on the nine-task average, with a mean gain of 15.6 percentage points. These results demonstrate the value of execution-improved software as a resource for acquiring new policies in this setting, although initial references remain better on two target tasks when averaged across runs.

So Kuroki, Yujin Tang

Sakana AI

Robot Policy Learning Robot Policy Adaptation Code Generation Policy Learning

Sep 17, 2026cs.AI

Dual-Axis Policy Optimization for LLM Agents: Bayesian Feedback Attribution and Trajectory Mass Normalization

Reinforcement learning for LLM agents involves two distinct optimization di- mensions: how environment feedback is exploited within a trajectory, and how complete trajectories are aggregated across a batch. We formulate these dimen- sions as Intra-Trajectory Feedback Attribution and Inter-Trajectory Objec- tive Aggregation, and introduce BATON (Bayesian Attribution and Trajectory Objective Normalization), a dual-axis policy optimization framework. BATON instantiates the first axis with Bayesian Feedback Attribution, which constructs a feedback-conditioned posterior over sampled actions, and the second with Trajec- tory Mass Normalization (TMN), which assigns equal optimization mass to com- plete trajectories. Experiments with GRPO and GiGPO on ALFWorld, WebShop, and SearchQA show that both axes provide independent gains and that their combi- nation consistently achieves the strongest overall performance across model scales.

Yingxuan Zhuang, Binhe Yu, Jingxiao Yang +6

Zhejiang University · University of Science and Technology of China · University of New South Wales

Group Relative Policy Optimization Policy Optimization Policy Learning RL for LLM Agents

Sep 16, 2026econ.EM

Stable Policy Learning

In evidence-based policymaking, typically one experimental sample is observed, then a learned policy recommendation is implemented at scale. Policies learned from the experimental data can perform well in expected welfare, yet random sampling in the experiment can produce recommendations with poor welfare outcomes. In this paper, we ask: how should policy learning algorithms balance expected welfare against sampling risk? Our main contribution is to show that algorithmic stability plays a central role in characterizing and navigating the tradeoff. Intuitively, if a policy learning algorithm's recommendation remains stable when one experimental unit is replaced, then that algorithm has limited sampling risk. We propose a method for policy learning called policy-vote bagging, which learns treatment decisions on many subsamples then averages their votes into treatment probabilities. Relative to using one subsample, averaging across subsamples preserves expected welfare and improves expected utility for a risk-averse researcher. We derive sharp bounds linking estimation accuracy, subsample size, and welfare variation, including an exact guarantee under CARA utility.

Harvey Barnhard, Giacomo Opocher, Rahul Singh

Harvard University · University of Mannheim

Algorithmic Stability Policy Learning

Sep 16, 2026cs.LG

A Convergence Framework for Deep $V$ -Learning: Error Propagation and Sharp Action-Gap Bounds

We establish convergence bounds for deep

V

-learning with horizon

H

. The algorithm fits a scalar value function to targets from executed transitions and selects actions using a predictive model and the value function. For current observed-successor targets with fresh true-kernel outcomes, the conditional mean is

\mathcal{T}^βV

, which averages over behavior-policy actions. The Bellman optimality update is

\mathcal{T} V

. We decompose the update error into six residuals: fitting, transition reuse, target construction, replay, action selection, and exploration. Under

L^s

concentrability, their

L^p

norms (

p=s/(s-1)

) control expected

L^1

policy loss. The bound explicitly weights residuals from only the last

H-1

update blocks, plus an initialization term for shorter runs. We quantify the cost of a shared sampling distribution across horizon levels. For statistical error bounds of order

n^{-ν}

, we derive optimal continuous allocations and an integer allocation whose objective is within a factor

2^ν

of the constrained optimum. A margin condition with exponent

α

gives action error of order

Λ^{1+α/p}

, where

Λ

combines network drift and score error; a one-step construction proves the exponent sharp. Bounds on the distance between frozen and optimal scores transfer an optimal-gap condition to frozen-iterate gap bounds while retaining the mass of optimal ties. Survival probabilities and coverage conditions at deployment yield bounds for policies selected with approximate scores. Separate spatial ReLU networks per horizon level give a conditional neural regression rate, and the finite-state case gives a log-free expected fit rate. These results give expected policy-loss consistency for the fixed-horizon generative-reset approximate-ERM procedure with exact action scores and provide an explicit residual-decay criterion for FIFO/interleaved SGD.

Yury Kolomeytsev

Faculty of Computational Mathematics and Cybernetics Lomonosov Moscow State University

Reinforcement Learning Policy Learning Model-Based RL

Sep 14, 2026cs.LG

Bellman Policy Optimization

Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models (LLMs). We introduce Bellman Policy Optimization (BPO), a critic-free method derived from Policy Mirror Descent (PMD). For autoregressive generation with terminal rewards, BPO uses the Bellman equations to reformulate PMD as a trajectory-level objective. The reformulation avoids estimating state values at intermediate states. We prove that it has the same unique optimal solution as the original PMD objective. We derive the practical BPO loss by approximating this objective. Its mismatch-correction weight is a smoothed ratio of complementary token probabilities. Experiments on mathematical reasoning benchmarks demonstrate the effectiveness of BPO.

Zhuoqing Song, Haotian Xu, Xikun Zhang +1

1Apodex US, Inc. · 2Princeton University

Reinforcement Learning RL for Language Model Reasoning Policy Optimization Critic-Free RL Reinforcement Learning with Verifiable Rewards Policy Learning+1

Sep 14, 2026cs.LG

MInTRL: Off-policy Intervention can boost On-policy RL

Reinforcement learning with verifiable rewards is typically performed on-policy, keeping training data close to the current policy but limiting learning to trajectories that the policy can discover itself. Off-policy methods such as supervised fine-tuning, on the other hand, can leverage external knowledge beyond the base model's capabilities, but may suffer from large distribution shift. The key challenge is thus to expand exploration without sacrificing learnability. In this work, we introduce Minimal Intervention Reinforcement Learning (MInTRL), which expands the exploration frontier through sparse, local interventions in otherwise on-policy rollouts. During generation, a judge-intervention policy periodically reviews the current policy's output, replaces erroneous suffixes with short corrections, and immediately returns control to the policy. During training, MInTRL adopts a sequence-level advantage-regression objective that eliminates the need for importance sampling. We show that sparse, local interventions can substantially improve coverage beyond finite-budget on-policy sampling while preserving the overall on-policy nature of the resulting trajectories. Across math and code benchmarks, MInTRL consistently outperforms standard on-policy and off-policy baselines. Ablations show that MInTRL remains effective with self-intervention and across different judge policies, while performance peaks at moderate intervention intensity, highlighting the importance of intervening minimally. These results establish minimal intervention as an effective paradigm for enhancing on-policy RL.

Mingyu Chen, Yefan Tao, Gerald Friedland +2

Amazon Web Services · Boston University

Reinforcement Learning Reinforcement Learning with Verifiable Rewards RL Post-Training Policy Learning Off-Policy RL

Sep 14, 2026eess.SY

Adaptive Agent Design

We consider an agent acting against a general non-Markovian environment. The agent maintains its agent states, but is free to choose a transition kernel across those states and optimize its state-feedback control policies. We study the bi-level agent design problem that optimizes the transition kernel and the policy it induces, given said kernel with offline data of observations and actions obtained via a behavioral policy. For general environments, we show that a soft

Q

-learning algorithm converges almost surely to the fixed point of a soft Bellman equation defined by the stationary averages that the behavioral policy and the chosen kernel induce, and we delineate what separates the resulting policy from an optimal one. In partially observed Markov decision problems, we analyze convergence properties of parametrized transition kernel design via zero-th order and Bayesian optimization techniques.

Raj Kiriti Velicheti, Subhonmesh Bose, Tamer Başar

Department of Electrical and Computer Engineering and the Coordinated Science Laboratory at the University of Illinois, Urbana-Champaign, Urbana, IL 61801

Reinforcement Learning Q-Learning Offline RL Partially Observable Markov Decision Processes Policy Learning

Sep 14, 2026cs.AI

SCQ: Stabilizing Conservative Q-Learning with Sigmoid-Bounded Entropy

Offline-to-online reinforcement learning reduces interaction cost for real-world robot learning but suffers from persistent value estimation instability. Existing methods address this through pessimistic regularization, lower-bound calibration, and architectural normalization, but an overlooked source of instability lies in the entropy formulation: the standard log-entropy term can become negative, destabilizing policy updates. We introduce SCQ (Sigmoid-Bounded Conservative Q-Learning), which replaces this term with a sigmoid-bounded formulation that stays strictly positive. SCQ retains conservative Q regularization and return-based lower-bound calibration, stabilizing policy optimization without sacrificing exploration. We evaluate SCQ on D4RL (Minari) benchmarks under both single-demonstration and standard dataset settings, as well as on simulation and real-world visual tasks. SCQ matches or exceeds baseline performance while exhibiting more stable training dynamics across state-based and visual benchmarks, and transfers to four real-robot platforms including manipulation, wheeled, quadruped, and humanoid systems. A direct clipping intervention that removes negative log-probability contributions, together with gradient-matched positive-score controls, indicates that positivity rather than a particular score shape alone drives much of the improvement. Project website: https://scq-rl.github.io.

Xiefeng Wu, Shu Zhang, Zhaojie Chu +1

School of Computer Science, Wuhan University, Wuhan, China · School of Electronic Information, Wuhan University, Wuhan, China · School of Internet, Anhui University, Hefei, China

Robotic RL Offline RL Maximum Entropy RL Policy Learning Offline-To-Online RL

Sep 14, 2026cs.LG

Robust Policy Optimization via Adversarial Importance Sampling

Significant progress has been made in safeguarding deep reinforcement learning (DRL) policies against input perturbations. Developing robust DRL involves three main stages: algorithm design, implementation, and evaluation. In this work, we identify and address a key limitation at each stage. First, we introduce Adversarial Importance Sampling (Advis), a method that uses importance sampling over trajectories from standard training to estimate and optimize verifiable worst-case returns. Advis satisfies three desirable criteria not jointly achieved by prior work: it requires no additional environment interactions, no auxiliary networks, and captures long-term robustness. Second, we introduce advrl, a modular PyTorch library that provides clean, single-file implementations of existing robustness methods and adversarial attacks, facilitating rapid prototyping and enabling reproducible and traceable evaluations. Third, we revisit evaluation under learned adversaries and show that optimal adversarial hyperparameters do not transfer across agents, which can lead to an overestimation of robustness when using a limited set of attacker configurations. Accordingly, we evaluate policies against a large and diverse set of attackers, using 6-14x more configurations than prior work. Finally, we evaluate our approach on continuous control environments, demonstrating its effectiveness relative to existing baselines. The code is available at: https://github.com/AmineAndam04/advrl

Amine Andam, Jamal Bentahar, Mustapha Hedabou

Mohammed VI Polytechnic University · Khalifa University · Concordia University

Robust RL Policy Learning

Sep 14, 2026cs.LG

A Unified and Constrained View of Regularization-Based Robust Reinforcement Learning

Regularization-based methods have become a standard approach for training Deep Reinforcement Learning policies against adversarial input perturbations. In this paper, we unify these methods by deriving new upper bounds on the performance gap between the nominal and worst-case policies. Each upper bound is expressed as an existing regularization objective plus a KL-divergence penalty between the nominal and worst-case policies, which further explains why adding a KL penalty improves robustness in practice. Building on these bounds, we formulate robust training as a constrained optimization problem, showing that existing methods correspond to the special case of a fixed Lagrange multiplier. We instead update the multiplier jointly with the policy to automatically tune the regularization weight. Finally, we conduct extensive adversarial evaluations across several continuous control tasks to validate our theoretical analysis.

Amine Andam, Jamal Bentahar, Mustapha Hedabou

Mohammed VI Polytechnic University · Khalifa University · Concordia University

KL-Regularized RL Adversarial Robustness Constrained RL Robust RL Policy Learning

Sep 12, 2026cs.LG

T1: Terminal Agent Reinforcement Learning for Long-Horizon Tasks

Agent usage is shifting toward long-horizon tasks such as coding and scientific discovery, among which terminal tasks are especially important. We introduce T1, a Mixture-of-Experts model of 122B total trained with reinforcement learning, operating a real shell in a cloud sandbox for up to 300+ tool-call turns per task, rewarded by executing each task's own verifier. We provide a comprehensive recipe: First, an aggressively warm-started to stabilize actor-critic training, with a dense process reward scoring trajectories by the absolute number of passing verifiers. Second, stable optimization through TITO construction, training on the exact sampled token identifiers with drift repair at turn boundaries, and rollout routing replay, recording the sampler's per-token expert choices at every MoE layer and replaying them during training. Third, fully out-of-distribution training corpus: isolated seeds and synthesized tasks disjoint from Terminal-Bench 2.1 ensures gains reflect genuine capability transfer over benchmark overfitting. Together, TITO and R3 cut the training-to-inference log-probability difference from 0.021 to 0.013, with exactly aligned zero token drift in the loss region. On Terminal-Bench 2.1, our post-train pipeline raises initial base model from 43.8% to T1 with 64.0% resolved. On Long-Horizon Terminal Bench, T1 reaches 27.9% and surpasses GPT-5.4 and GLM-5.1.

Junyao Yang, Yucheng Shi, Zhongzhi Li +4

Tencent Hy Foundation Model · National University of Singapore · University of Georgia +2

Terminal Agents Long-Horizon Agent Evaluation Reinforcement Learning with Verifiable Rewards Policy Learning RL for LLM Agents RL for Tool Use

Sep 11, 2026cs.LG

Granularity-Adaptive Credit Assignment for Long-Horizon LLM Agent Reinforcement Learning

Long-horizon language-model agents trained with reinforcement learning oftenreceive sparse outcome rewards that do not reveal which decisions along a tra-jectory deserve credit. Episode-level advantages provide coarse trajectory-widecredit, while step-level comparisons offer finer resolution with context-dependentestimation noise. We propose Granularity-Adaptive Credit Assignment (GACA),a critic-free method that adaptively mixes episode- and step-level credit for eachdecision during policy optimization. GACA normalizes the sampled response'smean per-token negative log-likelihood (NLL) within each trajectory and uses theresulting criticality score to determine the step-specific mixture. The computationreuses rollout log-probabilities without additional training rollouts or model eval-uations. Our analysis characterizes optimal score-dependent mixing and derivesconditions linking expected NLL to a lower bound on the preferred step-levelweight. Across ALFWorld and WebShop with 1.5B and 7B backbones, GACAachieves the highest reported mean success rates among the compared methods,while introducing negligible additional computation.

Taoran Liang, Yang Liu, Shang Luo +10

Nankai University · Peking University · Supply Chain Tech Team Y, JD.com +2

Credit Assignment in RL Step-Level Credit Assignment in RL Critic-Free RL Credit Assignment Policy Learning RL for LLM Agents

Sep 11, 2026cs.RO

IMLE-VLA: Fast Single-Step Action Generation for Vision-Language-Action Policies

Vision-language-action (VLA) policies leverage pretrained vision-language backbones to achieve strong cross-task generalization. A leading design couples this backbone with a dedicated continuous action head trained via diffusion or flow matching. However, such heads rely on iterative multi-step sampling, for example 10 Euler steps in

\pi_{0.5}

. This creates an inference bottleneck that produces stop-and-go movement in the robot and slower task completion. We introduce IMLE-VLA, which replaces the iterative action head with a single-step conditional generator trained via conditional Implicit Maximum Likelihood Estimation (cIMLE). The cIMLE objective promotes multimodal action coverage, avoiding the mode collapse of naive regression heads while eliminating multi-step sampling entirely. When IMLE-VLA is applied to

\pi_{0.5}

, it increases inference frequency 3.67x (55 Hz vs. 15 Hz), enabling up to 11x higher action throughput. On the 40-task LIBERO benchmark, IMLE-VLA achieves the highest average success rate (98.0%) among all baselines while leading in inference frequency. Under the test-time perturbations of LIBERO-plus, IMLE-VLA retains

\pi_{0.5}

's robustness while other baselines degrade sharply, confirming that the cIMLE head preserves generalization. Real-world experiments on a Franka Emika Panda across four tasks demonstrate smoother motion (2.2x to 3.0x lower jerk) and faster task completion, with IMLE-VLA outperforming

\pi_{0.5}

on every task and reducing average VLA inference time per episode by 3.9x to 6.6x. Videos and code are available at https://kianhk6.github.io/IMLE-VLA/

Kian Hosseinkhani (Simon Fraser University), Qinhe Peng (University of Pennsylvania), George Shramko (Simon Fraser University) +7

Simon Fraser University · University of Pennsylvania · Alberta Machine Intelligence Institute (Amii) +1

Efficient VLA Model Inference Vision-Language-Action Models Policy Learning

Sep 10, 2026cs.RO

DIA: Denoising Intermediate Advantage for Diffusion Policy Optimization

Diffusion-based robot policies have become widely used in robotic manipulation, where they are typically trained with behavior cloning. However, policies trained purely from demonstrations are limited by the quality and coverage of the available data. Reinforcement learning can further improve the performance of these pretrained policies through interaction. A common approach is to use policy-gradient methods that formulate diffusion-policy fine-tuning as an outer environment MDP together with an inner denoising MDP. However, existing methods typically assign the same environment-level credit to all denoising steps used to construct an action chunk, without distinguishing which intermediate decisions contributed most to the final return. We introduce Denoising Intermediate Advantage (DIA), a policy-gradient method that learns a value function over partially denoised actions and uses it to construct a denoising level advantage for each step of the generative process. DIA combines this inner credit signal with the standard environment-level PPO advantage, providing state-dependent credit throughout the denoising chain. Across Robomimic, FurnitureBench, Franka Kitchen, and D3IL, DIA consistently improves final performance over existing diffusion-policy fine-tuning methods. Beyond final reward, DIA reaches successful states more efficiently and can shift farther from the pretrained behavior distribution, enabling it to discover more effective and efficient task-level strategies and subtask sequences that baseline methods fail to reach.

Arjun Sohal, Yuchi Zhao, Miroslav Bogdanovic +1

University of Toronto, Toronto, ON, Canada · Vector Institute for Artificial Intelligence, Toronto, ON, Canada · Acceleration Consortium, University of Toronto, Toronto, ON, Canada +2

RL for Robotics Step-Level Credit Assignment in RL Diffusion Policy RL for Diffusion Models Policy Learning

Sep 8, 2026cs.RO

Proxy Policy Steering

Generalist robot policies carry broad manipulation priors from large-scale data, but specializing them to a new task remains the deployment bottleneck. This requires eliciting task-specific behavior from limited demonstrations without degrading their broad capabilities. We introduce Proxy Policy Steering (PPS), an inference-time adaptation method that resolves this challenge by training two lightweight proxy policies whose calibrated velocity-space difference steers the frozen base sampler. A reference proxy models the frozen base's behavior on target-task observations, and a task proxy, initialized from the reference, captures how this behavior changes under task supervision. Their difference forms a calibrated velocity-space residual that steers the frozen base sampler at every denoising step. We identify the conditions under which this residual isolates the change induced by task supervision, and validate them empirically. Because the base is never directly modified, its broad capabilities remain available at inference, including behaviors such as recovery from failure that the demonstrations themselves do not exercise. Adaptation requires only forward velocity predictions from the base, making PPS lightweight to train and applicable even without access to the base's parameters. On 8 real-world and 4 simulation manipulation tasks, PPS lifts the state-of-the-art pi 0.5 base policy by 53% absolute success rate on average, with zero-to-one gains on tasks the base never solves, while preserving the base's broad capabilities. PPS outperforms LoRA fine-tuning, from-scratch specialists, residual policies, and prior inference-time steering methods.

Chuanruo Ning, Tianrui Wang, Wei-Chiu Ma +1

Cornell University

Robot Policy Adaptation Robot Foundation Models Robotic Manipulation Robot Imitation Learning Policy Learning

Sep 7, 2026cs.AI

Mini-Batch Risk-Averse Deep Q-Learning: A Robot Navigation Case Study

We study the control of Markov decision processes in which the quality of a policy is evaluated by a dynamic, time-consistent Markov risk measure rather than by an expected discounted cost. The main obstacle to combining such measures with reinforcement learning is that a transition risk mapping depends on the transition kernel in a nonlinear way, and therefore cannot be estimated from a single observed transition. We remove this obstacle by employing mini-batch transition risk mappings: the mapping is applied to the empirical measure of

N

independent next-state samples, and the result is averaged. The resulting mapping is again coherent. However, as an expected value of a function of

N

next-state values, it admits an unbiased one-sample estimator. We embed this mapping into a double deep Q-network, analyze the two sources of estimation bias that arise, and obtain a risk-averse Q-learning method applicable to state spaces far beyond the reach of tabular schemes. The method is applied to an underwater robot navigation problem, in which a vehicle must visit collection points, gather stochastic information payloads, and deliver them at transmission points, while exposed at each step to the risk of destruction. A hierarchical decomposition delegates path execution to an exact graph search and confines learning to the high-level ``collect or transmit'' decision. A low-dimensional feature map, invariant under the symmetries of the problem, replaces the raw state--configuration encoding. In experiments on

300

held-out environments, the resulting policies transfer to instance sizes never seen in training, and already

N=2

reduces the upper semideviation of the outcome distribution while simultaneously improving its mean whenever the simulator is misspecified---an empirical counterpart of the duality between coherent risk measures and distributional robustness.

Aayush Patel, Andrzej Ruszczyński

Department of Management Science and Information Systems, Rutgers University

Deep Q-Learning Robot Navigation Risk-Sensitive RL Policy Learning

Sep 3, 2026cs.AI

DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training

Reinforcement Learning from Verifiable Rewards works well when a task has a programmatic checker, but most long-horizon agent domains have none. We work in the outcome-blind setting, where ground-truth success signals are not available. Multi-criteria rubrics are a popular way to supply such a reward; they are scored once per trajectory, but a single scalar is a poor signal across tens of steps. We propose DRACO: Distributing Rubric-based Advantage for Credit Optimization. It generates rubrics dynamically during training to track the policy's evolving capability, scores those rubrics once per completed trajectory, and redistributes that judgment over the steps responsible for annotated rubrics to produce differentiated per-step advantages in GRPO. The redistribution is closed-form and does not introduce any trained attribution module. On AppWorld, DRACO gains 15.9 points over the base model and 5.3 points over GRPO trained with a sparse ground-truth reward, despite not using any verifiers itself. On out-of-domain Tau-Bench, it gains 5.3 points over the base model even without a frontier judge, beating both ground-truth-reward training and other rubric-based training settings. The code for DRACO is available at https://github.com/IBM/draco.

Shubham Gandhi, Saurabh Goyal, Kiran Kate +1

1Carnegie Mellon University · 2IBM Research

Agentic RL Step-Level Credit Assignment in RL Rubric-Based RL RL Post-Training Policy Learning Fine-Grained Credit Assignment

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.LG

Headroom-Drift Replay: A Primitive for Principled Replay Control in GRPO

RL-based post-training for reasoning models is increasingly bottlenecked by repeated fresh rollout generation, particularly in agentic settings where environment interaction dominates wall-clock cost. Replay can reduce this burden by reusing past trajectories, but existing methods typically embed it within larger training pipelines involving exploration, experience restructuring, or mixed-policy optimization. This makes replay's own contribution difficult to isolate. We ask a focused question: how far can principled replay selection alone go? We introduce Headroom-Drift Replay, a group-level replay control primitive for GRPO that separates reuse into two decisions. Headroom ranks stored groups by remaining learning value, while Drift gates them by compatibility with the current policy. The fresh on-policy stream remains unchanged, and the method adds no auxiliary generation or training machinery. Across mathematical reasoning, multimodal reasoning, and Agentic Search benchmarks, this single intervention outperforms naive replay and matches or exceeds broader replay methods on Avg Mean@32. In Agentic Search, where environment interaction dominates cost, it delivers comparable quality at materially lower wall-clock time.

Hyun Bin Park, Du-Seong Chang

Department of Artificial Intelligence, Sogang University, Seoul, Republic of Korea

RL for Language Model Reasoning Group Relative Policy Optimization Experience Replay RL Post-Training Policy Learning RL for LLM Agents

Sep 3, 2026cs.LG

Multi-step Proximal Policy Improvement in Offline Reinforcement Learning

Offline reinforcement learning (RL) must reconcile two competing requirements: policy updates should stay near dataset-supported actions to keep value estimates reliable, yet meaningful gains often require moving beyond the behavior distribution. We develop a geometric view of offline actor updates by modeling policies as a probability manifold endowed with a chosen metric geometry. Under this lens, a broad class of offline actor objectives can be interpreted as a single proximal policy improvement step (SPI), i.e., an implicit discretization of a manifold gradient flow induced by a critic-defined energy. Building on this insight, we propose multi-step proximal policy improvement (MPI), a plug-in refinement mechanism that composes sequential re-centered proximal steps. MPI enables controlled policy improvement beyond dataset support while retaining proximal control at each refinement. The framework accommodates multiple policy geometries and admits practical instantiations for deterministic and diagonal-Gaussian policies. Experiments on D4RL benchmarks show that small numbers of MPI refinements improve strong offline baselines, including TD3+BC, ReBRAC, and IQL, on many tasks. Focused diagnostics further distinguish re-centered refinement from fixed-objective update scheduling and characterize limitations under critic error.

Soohyun Choi, Seonvin Cho, Songnam Hong

Department of Electronic Engineering, Hanyang University, Seoul, Republic of Korea

Offline RL Policy Learning

Sep 2, 2026cs.LG

Tail-Likelihood Reinforcement Learning

Reinforcement learning typically optimizes average reward. For generative policies, the average can hide an important distinction: two policies can achieve the same mean reward while having very different chances of producing a rare but high-reward rollout. This matters as sampling increases during training and inference, since its benefit depends on retaining probability mass on high-reward outcomes. We propose to optimize this coverage directly. Rather than considering only expected reward, we consider all of its upper tails: for each reward threshold, how likely is the policy to exceed it? This turns a continuous reward into a family of binary success events. We introduce Tail-Likelihood Reinforcement Learning (TailRL), which maximizes the log-probability of exceeding a randomly chosen reward threshold. Its gradient gives more weight to rare, high-reward rollouts and can be interpreted as a mixture of Best-of-k gradients. TailRL requires only a simple modification to the advantage function, making it compatible with existing reinforcement learning pipelines. Across object localization, maze navigation, GUI grounding, and code optimization, TailRL leverages rare high-reward training samples to avoid suboptimal solutions and yields models that benefit more from additional samples at inference time.

Shrinivas Ramasubramanian, Daman Arora, Fahim Tajwar +11

Carnegie Mellon University · Together AI · University of California, Berkeley +2

Risk-Sensitive RL RL for Generative Models Policy Learning