Reinforcement_Learning
-
RobustRLlib: A Unified Library and Benchmark for Robust Reinforcement Learning Algorithms
2026-10-23 17:57:22Xi Chen, Fengqi Hu, Zeyu Shen, Jiaying Xu, Shangding Gu, Pan Xu, Anqi Liu, Laixi ShiAbstract
Robust reinforcement learning (RL) is critical for reliable real-world deployment, where policies may fail under diverse shifts and disruptions. Existing robust RL algorithms offer a rich set of mechanisms and principles, yet two major barriers limit their broader potential: limited algorithm reusability and an unclear empirical capability boundary. We introduce RobustRLlib, an algorithm-centric library and benchmark that addresses both challenges. It integrates 17 representative robust online, offline, and safe RL algorithms under a unified interface for substantially improved reusability, together with an enhanced evaluation platform spanning diverse real-world shifts and advanced humanoid and vision-language-action tasks. Systematic evaluation over 273 perturbation configurations provides an up-to-date capability map of current robust RL algorithms, revealing where they remain effective and where substantial robustness gaps persist. As an open-source, user-friendly platform, RobustRLlib offers a reproducible and extensible codebase for applying existing methods to new problems and developing new robust RL algorithms.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32971
PDF
-
RynnUnify: Towards Unified Embodied Brain
2026-10-23 17:54:07Minghao Zhu, Bohan Hou, Kun Xiang, Zixuan Wang, Yinzhou Tang, Yifei Yang, Zhikai Wang, Tianyi Zhang, Kehan Li, Haoyu Zhao, Siteng Huang, Leigang Qu, Sicong Leng, Dongchi Huang, Mingxiu Chen, Shijian Lu, Yue Wang, Xiaodan Liang, Zhenyu Zeng, Deli Zhao, Chengju Liu, Jianfei Yang, Xin LiAbstract
We introduce RynnUnify, a 16B unified embodied model that combines strong embodied understanding with long-horizon visual planning through interleaved video--language reasoning. With capabilities spanning spatial reasoning, physical prediction, and grounding, RynnUnify connects an understanding of the physical world with visual anticipation of how it evolves across successive subtasks. Building on these capabilities, we further develop an agentic extension through reinforcement learning to selectively use video predictions for reasoning about unobserved states and interaction outcomes. We also develop a world-action extension of the full RynnUnify model, translating its understanding and prediction capabilities into robot control. To evaluate these capabilities, we introduce Rynn-VP and Rynn-GEU for long-horizon visual planning and anticipatory reasoning. Experiments show that RynnUnify achieves the highest average embodied understanding score among the evaluated unified multimodal models and improves the Rynn-VP overall score by 9.75 points over Cosmos 3 Nano. Its agentic extension improves the Rynn-GEU average by 6.0 points over direct reasoning, while the world-action extension achieves 51.5\% success on 20 selected composite RoboCasa365 tasks and 46.7\% average success across three real-robot tasks. Together, these results establish RynnUnify as a shared foundation for embodied understanding, visual planning, and robot control, taking a step toward a unified embodied brain.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32965
PDF
-
Where Rounding Happens Matters: Rounding-Location Alignment Stabilizes LLM Reinforcement Learning
2026-10-23 17:45:03Yijing Ke, Tianle Cai, Ruosong WangAbstract
Reinforcement learning for large language models often uses separate systems for inference and training. Even with identical weights and BF16 tensors, implementation differences can make an intended on-policy update off-policy and destabilize training. Existing remedies align the entire inference and training stacks, change global precision, or modify the learning algorithm. We ask whether aligning selected parts of the execution stack can stabilize BF16 RL while other implementation differences remain. We align training-side rounding locations with inference in three regions---RoPE, SwiGLU, and residual RMSNorm---by retaining selected intermediates in FP32 until their BF16 outputs. Across four Qwen-family models, aligning all three regions prevents the collapse observed under the default BF16 baseline; RoPE-path alignment alone also suffices in the studied runs, despite remaining numerical discrepancies. The interventions keep the inference engine, matrix-multiplication and attention kernels, and learning objective unchanged.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32951
PDF
-
HoloBoxing: A Simple Recipe for Humanoid Boxing with 7 Minutes of Single-GPU Training
2026-10-23 17:11:28Xihan Ma, Bo Zhang, Boxiong Li, Yuhan Bao, Ziyue Liu, Yucheng WangAbstract
HoloBoxing presents a simple recipe for learning humanoid boxing by building on a pretrained whole-body controller. The recipe separates reusable motor skills from task decisions: a compact planner requests future body-point targets, while the frozen controller handles joint-level execution. A learned decoder maps 32-dimensional actions to structured target windows, and measured-state imitation initializes the planner before reinforcement learning. Task learning then refines interaction behavior while reusing the decoder and controller, avoiding the need to relearn whole-body control. This combination keeps task optimization focused on a compact upper policy and supplies an initial behavior before reward-driven exploration. Starting from that available controller, the recipe accounts for about seven minutes of additional training, including reused representation fitting and a single-GPU task run. This excludes controller pretraining, data preparation, evaluation, and search. Simulated boxing experiments and component controls evaluate the resulting behavior and the roles of imitation initialization and the target representation.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32909
PDF
-
NgramRL: Accelerating RL Post-Training for Structured Generation via Rollout-Aware N-gram Speculative Decoding
2026-10-23 16:47:29Nilesh Prasad Pandey, Jangseon Park, Jason Kong, Onat Gungor, Tajana RosingAbstract
Reinforcement learning (RL) post-training improves the reasoning capabilities of large language models, but its cost is dominated by autoregressive rollout generation, which can exceed 70% of training time. Existing speculative-decoding methods are poorly suited to this setting: model-based drafters consume scarce GPU memory and may become misaligned as the policy evolves, while existing model-free methods do not explicitly exploit repetition across training epochs. We observe that recent rollouts of the same prompt provide a high-quality, self-updating draft source. Based on this insight, we propose NgramRL, a per-prompt n-gram drafter constructed from bounded recent rollout history. NgramRL requires no auxiliary model, drafter training, or persistent GPU-resident drafter state, while target-model verification preserves the rollout distribution. To hide CPU cache lookup and scheduling, we further introduce an adaptive fan-out prefetcher that predicts likely verification outcomes (j, b), consisting of the accepted draft length j and target-emitted continuation token b. A causal budget allocator concentrates prefetches on high-probability outcomes, while uncovered outcomes fall back to sequential lookup without changing committed tokens. Across two model families and three reasoning and structured-generation workloads, NgramRL commits 3.60-7.77 tokens per target-model forward in the warm-cache regime, up to 1.99x more than SuffixDecoding and 7.0x more than prompt-lookup decoding. On Spider, the prefetcher covers 96.6% of Qwen3-8B outcomes at a candidate budget of 16 and improves measured rollout throughput by 1.38x over synchronous NgramRL, approaching its 1.40x overlap ceiling.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32881
PDF
-
Regularised RL improves sim-to-real transfer
2026-10-23 16:28:10Danilo Vucetic, Mohamad H. Danesh, Florence Clerc, Hsiu-Chin Lin, Jonathan TremblayAbstract
Regularised reinforcement learning (RL) has provable robustness guarantees that should ease the transfer of policies from simulation to robots, known as sim-to-real. Yet, weakly regularised algorithms like Proximal Policy Optimisation (PPO) remain the de facto choice in sim-to-real pipelines, while the practical benefits of regularisation for transfer remain poorly characterised. In this work, we study the sim-to-real performance of commonly used RL algorithms, with the aim of characterising the effects of regularisation on robotic performance. To that end, we derive Soft MellowMax Actor-Critic (SMM-AC), a novel regularised RL algorithm based on the soft mellowmax value function, which carries strong robustness guarantees. On MuJoCo continuous control benchmarks, SMM-AC is competitive with baselines like Soft Actor-Critic (SAC). On real robots, SMM-AC achieves higher return, while exhibiting greater robustness to perturbations in observations, the environment, and dynamics than baselines. Together, these results demonstrate that regularised RL improves real-world policy deployment through better robustness to the sim-to-real gap, and changing real-world conditions.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32855
PDF
-
Predictive Long-Horizon Progress Modeling with Action-Aware Quantification
2026-10-23 15:25:06Tianchi Xu, Yalin Zheng, Yangxin Zhu, Ruonan Ren, Zheng Fu, Yunlong Wang, Yuanxin ZhongAbstract
Quantitative and dense task progress feedback is essential for robotic manipulation across data curation, policy evaluation, and reinforcement learning. Existing progress annotations often rely on sparse keyframe markings derived from subjective manual judgments, limiting their scalability and cross-task consistency, particularly in long-horizon manipulation. To overcome these challenges, we propose action-aware quantification, a unified progress specification that decomposes long-horizon trajectories into object-level subtasks via a vision-language model (VLM) and assigns weighted action credits to capture fine-grained execution dynamics. Using VLM auto-labeling and human calibration, we construct an 800+ hour training video dataset and LOOM, a long-horizon progress benchmark containing 19,618 human-calibrated evaluation frames. To model these dense progress dynamics, we introduce NimCritic, a predictive progress critic trained via Joint Video-Progress Learning. We further incorporate Historical Progress Perturbation to mitigate sequential drift, and design a contrastive Video-Progress Alignment objective that bridges asymmetric expert streams by mutually grounding intermediate visual dynamics and denoised progress trajectories. NimCritic matches or outperforms state-of-the-art process reward models (PRMs) on the established Video Progress Benchmark (VPB) and LOOM, surpassing GVL (GPT-6-Sol) on VPB by about 65% in terminal success accuracy (TSA). On the large-scale RoboArena policy evaluation benchmark, it also retrieves human-selected successful rollouts with about 97% overall recall. These results establish action-aware quantification as a scalable foundation for dense progress feedback in robotic manipulation.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32776
PDF
-
Optimization, Optimal Control, and Reinforcement Learning for Robotics
Abstract
These lecture notes provide an introduction to optimization, optimal control, and reinforcement learning for robotics, emphasizing their mathematical connections and practical relevance. They present established material compiled and edited by Yonghyeon Lee, rather than new research results. The notes were developed through a human and AI collaborative writing process, with AI assistance in drafting, organizing, and refining the exposition. This first version is a preliminary collection of notes, not yet a fully developed textbook, and currently contains no figures or graphs. The notes may be revised, corrected, and expanded in future versions.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32538
PDF
-
MusicTwins: Training Audio-Language Models to Listen like Their Persona
2026-10-23 11:11:45Milan Liessens Dujardin, Kevin MiaoAbstract
People differ in how they perceive music, what associations they make with it, what they like about it, and how they respond to it. Computational models intended to simulate people in environments that include music should account for these differences. Although LLMs have been shown to reproduce aspects of human behavior, whether audio-language models (ALMs) can reproduce a specific person's responses to textual and musical stimuli remains underexplored. We introduce MusicTwins, a framework that combines persona summaries, personal memory, and post-training to generate person-specific responses. We construct persona representations from human writings, capturing background, views, personality, writing style, and a musical profile, and post-train Audio Flamingo Next through supervised fine-tuning and reinforcement learning. Evaluation on ten held-out personas shows notable improvements for Audio Flamingo Next in persona fidelity, including personality, writing style, and responses to music, supporting MusicTwins as a foundation for training audio-language models to listen like their persona.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32495
PDF
-
Accordion: Elastic-Depth Inference via Fixed-Prefix Layer Looping
2026-10-23 11:07:45Yudong Zhang, Wei Ding, Kaili Zheng, Yuchen Ding, Zhenyu ZhangAbstract
Large language models serve applications with diverse inference budgets, but training and maintaining separate models for each budget incurs substantial costs. Recent Looped Transformers enable flexible inference computation through parameter reuse, but maintaining strong performance across different inference budgets remains an open challenge. We introduce Accordion, a mixture-of-experts-based model that supports five inference budgets through fixed-prefix layer looping. Each budget follows a pre-defined execution path, while the minimum budget retains the unlooped model. Joint pretraining samples one budget per micro-batch and optimizes its next-token prediction loss without additional cross-budget consistency objectives. We further preserve these inference budgets through supervised fine-tuning and reinforcement learning, where parameter-group merging and budget-matched multi-teacher on-policy distillation integrate three domain experts into one model. After 3.6T tokens of pretraining, increasing inference depth improves Accordion-6B-A0.5B's five-category macro-average across 37 tasks by 3.01 percentage points, with the middle budget capturing 89% of this gain using half as many additional layer executions. The final model surpasses each domain expert on its corresponding task-and-budget average, while achieving strong instruction-following performance across all budgets and retaining depth-dependent gains in mathematical reasoning and code generation.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32492
PDF
-
$R^2E:$ A Role-driven Reward Evolutionary Framework for Automated Reward Function Design
2026-10-23 10:06:40Shouhao Chang, Xuan Liu, Hongye Zhu, Xinning Chen, Shigeng ZhangAbstract
Designing effective reward functions is a critical challenge in Reinforcement Learning (RL), which traditionally requires costly manual trial-and-error. While Large Language Model (LLM)-based methods have shown promise in reward design, undifferentiated sampling and limited use of historical feedback often lead to insufficient exploration of the reward space and optimization instability. To address these challenges, we propose $R^2E$, a Role-driven Reward Evolutionary framework for automated reward function design that structures the process through explicit role specialization. $R^2E$ employs multiple LLM roles with complementary objectives: an Explorer that promotes novelty to expand the search space, a Guardian that performs conservative refinements to improve training stability, and an Evolver that recombines high-performing rewards via crossover and mutation to integrate effective structures. A dynamic scheduling strategy coordinates these roles, progressively shifting the search from broad exploration to focused exploitation. To mitigate optimization instability and provide a medium for cross-role collaboration, $R^2E$ incorporates a Global Elite Pool that retains the best-performing rewards to guide subsequent generations, thereby enabling coordinated interaction among roles and ensuring a reliable and consistent refinement process. Extensive experiments on multiple robotic tasks demonstrate the effectiveness of our method.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32439
PDF
-
Profile-Driven User Simulation via Reinforcement Learning with Hierarchical Rewards for Task-Oriented Dialogues
2026-10-23 10:01:41Juan Li, Fangshu Chang, Xinyi Jiang, Ling Huang, Yunzhang ChenAbstract
Building Task-Oriented Dialogue (TOD) systems for high-value scenarios such as financial marketing requires large volumes of domain-specific dialogue data, which privacy regulations, annotation costs, and the need for expert labor make hard to obtain. User simulation offers a remedy, but existing methods target general-domain dialogue and exhibit two key limitations: insufficient modeling of domain-specific business processes and compliance constraints, and inadequate capture of the colloquial, spoken-language character of voice interactions. We propose ProfileSim, a profile-driven user simulation framework for financial marketing outbound calls. It jointly models user profiles---demographics and Big Five personality traits---and business profiles---intents reverse-engineered from real call logs---enabling fine-grained control over simulated behavior. The base model is fine-tuned on real voice dialogue data and refined by Group Relative Policy Optimization (GRPO) with hierarchical rewards at the dialogue and utterance levels. Paired with a customer service agent integrating Standard Operating Procedure (SOP) constraints and Retrieval-Augmented Generation (RAG), ProfileSim achieves the highest colloquial naturalness, the strongest user profile adherence and the lowest agent score on long-tail intents among all evaluated user generators, while producing substantially shorter user utterances at comparable business compliance. The generated dialogue data is designed to support intent recognition training, agent evaluation, and rapid cold-start deployment.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32435
PDF
-
When to Stop Penalizing Actions? Martingale GRPO for VLA Reinforcement Learning
2026-10-23 09:56:27Zongyao Zhao, Cheng-Jie Zhao, Ruihang Jiang, Bing Zheng, Zhaolin Wang, Yuanwei LiuAbstract
A failed vision-language-action (VLA) rollout does not imply that every action contributed to failure. Yet in group-relative policy optimization (GRPO), a single trajectory-level advantage is assigned to all actions in a rollout. Potentially useful recovery attempts can thus be penalized alongside the errors they aim to correct. Ideal stopping conditions for failure feedback are characterized using success-probability and policy-score martingales. At these stops, the removed gradient contribution has zero conditional expectation given the retained information. Truncation therefore preserves the expectation of a reference GRPO group-gradient estimator and can reduce its covariance. One such stop occurs when the conditional probability of success is zero, motivating error and recovery cues for selecting candidate feedback boundaries. Martingale GRPO (MART-GRPO) is introduced with a vision-language model (VLM) for event-guided truncation. The VLM is fine-tuned on our annotated training set using low-rank adaptation (LoRA). The last visible error without observed recovery is localized by the frozen VLM in complete rollout videos. Failure feedback is retained through the selected error interval and removed afterward, while successful-rollout loss terms and original loss normalization are preserved. For retrospective boundaries, changes in expectation and variance are analyzed to assess departures from ideal stopping. Experiments with OpenVLA-OFT on RoboTwin 2.0 show that MART-GRPO improves success rates over full-trajectory GRPO across multiple tasks.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32431
PDF
-
From Visual Affinity to Learning Credit in Multimodal Reinforcement Learning
2026-10-23 08:22:16Yuhang Han, Dingning Liu, Yujie Chen, Yaojie Zhang, Jingwei Song, Zequn Yang, Xuanming Shang, Li Ding, Zhijiang Guo, Linfeng ZhangAbstract
Reinforcement learning with verifiable rewards (RLVR) improves multimodal reasoning, but uniformly assigning a single response-level advantage to all tokens overlooks the distinct roles of visual observations and subsequent reasoning. Visual affinity alone is insufficient: predictive states may align with both image and prompt-context representations, and upweighting valid visual observations can amplify their negative-advantage updates when later reasoning fails. To address this issue, we propose \textbf{Outcome-Conditioned Visual Credit (OVC)}, a plug-in framework that separates visual-association measurement from token-level credit assignment and is adaptable to policy-optimization methods with uniform response-level advantage allocation. Relative Visual Encoding constructs a context-relative visual coordinate by contrasting each predictive state's affinity to image and prompt-context representations. Outcome-conditioned Evidence Calibration combines this coordinate with uncertainty-scaled predictive features and calibrates their contributions to outcome-discriminative evidence through mean contrasts between positive- and negative-advantage responses, scaled by regularized within-side variation. Visual association thus informs credit through observed outcome differences rather than directly determining update priority. Normalized Credit Allocation conditions the calibrated evidence on the response's advantage sign to produce positive token weights with unit global mean. These detached weights preserve advantage signs while modulating individual token-gradient contributions. OVC reuses existing rollouts without auxiliary reward models, process annotations, or inference-time modules. Instantiated through GSPO-token, OVC retains sequence-level clipping and consistently improves six-benchmark mean accuracy over GSPO across three models, with gains of 1.86-2.39 percentage points.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32324
PDF
-
Spatial Heuristic Strategy based Diffusion Reinforcement Learning for Online 3D Bin Packing
2026-10-23 06:49:30Tong Li, Qingyang XuAbstract
Traditional heuristic methods have accumulated mature experience in core objectives, such as space utilization and layout compactness of 3D bin packing, relying on artificially designed rule-based logic. However, they exhibit obvious limitations in generalization ability and adaptability to dynamic scenarios. Diffusion reinforcement learning, which integrates the spatial search capability of reinforcement learning and the data distribution modeling capability of diffusion learning, has achieved certain progress in action distribution modeling and decision compliance constraints. Nevertheless, there remain issues including insufficient perception of the spatial topological structure in 3D bin packing and limited layout optimization capability. To address these problems, we propose a diffusion reinforcement learning method integrated with spatial heuristic strategies, which combines classical bin packing heuristic rules with a diffusion reinforcement learning framework. Specifically, it compresses the action space through heuristic action pre-screening, introduces a graph neural network to model the spatial structure, and designs a composite reward function incorporating flatness to suppress spatial fragmentation, thereby realizing the organic integration of prior knowledge and adaptive optimization. Experimental results demonstrate that the pro
Abstract (translated)
URL
https://arxiv.org/abs/2610.32241
PDF
-
Financial Context-Rich Strategy for Actor-Critic based Agents: A Deep Reinforcement Learning Approach for Stock Trading
2026-10-23 06:46:09Hrishikesh Singh, Aarti Sharma, Millie PantAbstract
In modern financial markets, more than 3/4th of the large volume trade is executed by algorithmic trading based on advanced trading strategies to attain higher precision and speed than traditional human traders. The conventional strategies focused on price-based indicators are inadequate in capturing the market dynamism arising from real-world events such as social sentiments, political instability, or sudden economic shifts. With the rise of Deep Reinforcement Learning (DRL) and data availability, there is a requirement for more efficient algorithms that can incorporate these external factors in trading strategies. DRL has shown potential in simulating real-world scenarios and in simplifying complex decision-making since DRL agents can automate the trading process i.e. executing trades in a manner comparable to human traders. They have shown reliable performance since they can update their policies based on risk and quantitative factors. In this paper, we propose a DRL approach based on partial and deep-infusion of sentimental analysis. It features distinct policy designs of actor-critic-based agents: only policy optimization (A2C & PPO) and policy optimization along with Q-learning principles (DDPG, TD3 & SAC). Our findings reveal, that the agents based on policy optimization combined with Q-learning, significantly and consistently demonstrate improved performance upon sentiment deep-infusion in profit maximization as well as risk-averse trading.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32237
PDF
-
ACWM-RM: A Reference-Free Reward Model from Complementary Mismatch Preferences
2026-10-23 06:27:26Shibo Hong, Siyu Ren, FengJiao Chen, Zhongyuan Peng, Wei Shi, Yixin CaoAbstract
Reinforcement learning has gained attention for improving action-conditioned world models (ACWMs), as it enables reward-driven optimization of visual quality (VQ) and action-following fidelity (AF) beyond flow-matching optimization. Realizing this potential requires reward signals that reliably rank videos by both VQ and AF. However, existing rewards capture only one dimension: metric-based rewards assess VQ but require ground-truth videos and do not directly assess AF, while inverse dynamics model (IDM) rewards assess AF without references but do not guarantee VQ. Moreover, IDMs, typically trained on real videos, may provide unreliable rewards for generated videos containing visual artifacts and physical inconsistencies. In this work, we introduce the Action and Video Mismatch Benchmark (AVM-Eval), comprising 67.2k pairs that evaluate how reliably IDMs assess AF on real and generated videos. We then propose ACWM-RM, a reward model that jointly assesses VQ and AF without ground-truth videos during reward computation. To train it, we first construct the Action and Video Mismatch Preference Dataset (AVM-Pref) with 506.3k pairs providing complementary supervision: action-mismatch pairs supervise AF, while video-mismatch pairs supervise VQ and expose the model to generated videos. We then pretrain an action-conditioned predictor on real videos to learn forward dynamics before jointly training it with a patch-weighted score head on AVM-Pref to transfer these dynamics to AF assessment on generated videos. Evaluation on AVM-Eval reveals the limited reliability of IDMs on generated videos, while downstream RL experiments show that ACWM-RM outperforms reward baselines in both criteria.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32222
PDF
-
SOPO: Socratic Guided Policy Optimization for Span-Level Hallucination Detection
2026-10-23 06:08:17Jiawei Dong, Zilong Bai, Pengtian zhu, Peng ZhouAbstract
Large Language Models (LLMs) have gained widespread adoption in various natural language processing tasks, but suffer from hallucination issues where they generate unfaithful or inconsistent content. While Reinforcement Learning with Verifiable Rewards (RLVR) has shown promising applications in reasoning tasks, its application to span-level hallucination detection is challenged by localization ambiguity, reward sparsity on hard examples, and uncontrolled length inflation. To address these issues, we introduce Socratic Guided Policy Optimization (SOPO), a multi-turn RL framework for robust hallucination detection. The key innovation of SOPO is a Socratic feedback mechanism, where a teacher model guides the reasoning trajectories during rollout by providing indirect hints inferred from the ground truth. This teacher-guided approach enables the model to generate more effective reasoning trajectories for hard examples, thus alleviating reward sparsity. Furthermore, we incorporate Turn-based Advantage Scaling to penalize verbosity bias, encouraging the model to develop efficient and autonomous reasoning capabilities. Empirical results on RAGTruth demonstrate that SOPO (4B) achieves state-of-the-art performance, surpassing leading proprietary models (e.g., GPT-5) and standard RL baselines, while obtaining higher reported mean scores than GRPO on four out-of-domain benchmarks, with mixed evidence against CAPO. We release our training and evaluation code alongside the 1.7B and 4B SOPO models at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32212
PDF
-
Dual-Adaptive Z-Score PPO: Dynamic Clipping Using Action Confidence and Normalized Advantage Magnitude
2026-10-23 05:40:26Taha M. Mahmoud, Naima KaabouchAbstract
Proximal Policy Optimization (PPO) remains a popular on-policy reinforcement learning algorithm owing to its stability and ease of implementation. Its clipping threshold, however, is fixed, so every training sample receives the same trust-region width regardless of how reliable the corresponding policy update is. This work examines adaptive-clipping variants of PPO in which the constant clipping parameter is replaced by sample-dependent thresholds computed from policy and return statistics. The central method, Dual-Adaptive Z-Score PPO (PPO-DAZ3), sets the clipping interval for each sample from two signals: action confidence, given by the probability of the selected action under the current policy, and update confidence, given by the normalized advantage magnitude. The rest of the standard PPO training pipeline is left unchanged. Evaluation on three classic control benchmarks (LunarLander-v3, CartPole-v1, and Acrobot-v1) indicates that adaptive clipping improves learning in some settings but not all. PPO-DAZ3 performs best on LunarLander-v3, the smooth advantage-based variant PPO-AZ3S leads on Acrobot-v1, and standard PPO stays competitive on CartPole-v1, suggesting that adaptive clipping is most beneficial on harder control tasks.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32194
PDF
-
$Psi$-0.6: Adaptive Streams of Thought in a Unified Vision-Language World Model
2026-10-23 05:24:02Khai Loong Aw, Klemen Kotar, Baihan Zhang, Yifan Liu, Lilian Chen, Atlas Kazemian, Rahul Venkatesh, Daniel YaminsAbstract
Humans reason about the physical world by flexibly combining different forms of inference as a problem unfolds. Achieving this flexibility is a major open challenge in AI: capabilities often remain separated across specialized models, limiting how they can be composed and jointly adapted to a task. We introduce $\Psi$-0.6, a unified vision-language world model that adaptively composes perception, language, geometry, and simulation within a shared reasoning process we call streams of thought. These streams interleave images, text, and explicit representations of geometry and motion, all of which the model can generate and interpret through a shared token interface. $\Psi$-0.6 surpasses leading open-weight models zero-shot on multi-view spatial reasoning, with further gains from reasoning through geometric intermediates. It also supports counterfactual simulation: generating and interpreting imagined scenes to determine, for example, whether moving an object would obstruct a pathway. Reinforcement learning from task feedback jointly optimizes the selection, generation, and use of intermediate representations, improving visual question-answering accuracy with gains that transfer to unseen benchmarks. $\Psi$-0.6 offers a path toward adaptive physical reasoning -- not only predicting the physical world but also choosing the representations through which to reason.
Abstract (translated)
URL
https://arxiv.org/abs/2610.32177
PDF
