DualSQL is a Text-to-SQL system that uses two agents sharing a single model backbone, enabling joint optimization via multi-agent reinforcement learning. The approach incorporates three database access tools for multi-step reasoning, rollout guardrails to stabilize training, and a new SQL correctness metric called robust execution match (REX). Trained on only 3,755 examples, DualSQL-4B reaches 68.0% execution accuracy on the BIRD dev set, while DualSQL-8B achieves 71.1%, surpassing prior state‑of‑the‑art single‑model solutions with 32B parameters.
By Shijie Chen, Yu Gan, Yeounoh Chung, Jiani Zhang, Quannan Li, Sravan Babu Bodapati, Cody J. Greer, Yu Su, Fatma Ozcan
The paper introduces a robust variant of Monte Carlo Tree Search that addresses ambiguities in transition dynamics and reward distributions, bridging the gap between simulation-based planning and real-world deployment. It incorporates a robust power mean backup operator and exploration bonuses to guarantee finite-sample convergence at every node, achieving an ≠O(n−1/2) convergence rate for root value estimation comparable to standard MCTS. Empirical results demonstrate robust performance in planning tasks even under significant model mismatches.
By Tuan Dam, Kishan Panaganti, Brahim Driss, Adam Wierman
CoRe-MARL is a cooperative multi-agent reinforcement learning framework designed for decentralized relief distribution networks. It models each local center as an agent in a Dec-POMDP, using a recurrent network to learn redistribution policies that reduce service gaps and improve the worst-served region. Experiments show that recurrent MAPPO outperforms independent PPO and heuristic baselines, maintaining competitive network-wide service while adapting to evolving supply and demand dynamics.
By Naimur Rahman Chowdhury, Shatabdi Sen Prapti, Md. Salehin Seyam, Limon Bin Hossain
RLLBC-Lib is an educational code library designed to lower the entry barrier for students learning reinforcement learning (RL) in the context of learning-based control. It offers a comprehensive collection of tabular RL methods to reinforce theoretical foundations, followed by a deep RL library that mirrors the same design principles to highlight parallels between simple and state‑of‑the‑art approaches. The library also includes implementations that illustrate core RL principles, contrast RL with other learning‑based control methods, and serve as a foundation for programming assignments with automated grading.
By Bernd Frauenknecht, Emma Cramer, Artur Eisele, Paul Kruse, Lukas Kesper, Jonas Hertrampf, Ramil Sabirov, Jyotirmaya Patra, Johannes Berger, Paul Brunzema, Friedrich Solowjow, Sebastian Trimpe
The paper introduces ASTRIL-MPC, a language‑guided neural model predictive control framework that enables articulated tracked robots to navigate complex, contact‑rich urban environments such as stairwells and cluttered interiors. By combining a learned kinematics model that predicts short‑horizon state changes, an optimization‑based planner with multi‑objective costs, and a large language model that safely updates control weights, the system achieves up to 71% better traversal quality than non‑adaptive NMPC and 67% better than a PPO baseline, while eliminating collision impacts during descent. Real‑robot trials over four indoor obstacles confirm the method’s transferability to physical contact‑rich traversal.
By Zhenfeng Gan, Yanbo Chen, Lirong Che, Yongyi Ma, Rongkai Zhu, Xueqian Wang
The paper introduces Variational Graph-to-Scheduler (VG2S), a framework that applies variational inference to the Job Shop Scheduling Problem (JSSP). By decoupling representation learning from policy optimization using a variational graph encoder and an ELBO-based objective, VG2S improves training stability and robustness to hyperparameter changes. Experiments show that VG2S outperforms state‑of‑the‑art deep reinforcement learning baselines and traditional dispatching rules, especially on large‑scale benchmark instances such as DMU and SWV.
By Seung Heon Oh, Jiwon Baek, Hyunjin Oh, Kiyoung Cho, Heechang Yoon, Jong Hun Woo
AuthorMix is a lightweight, modular framework for authorship style transfer that uses layer‑wise adapter mixing. It trains individual style‑specific LoRA adapters on a small set of high‑resource authors, enabling rapid adaptation to new target styles with only a few examples. The method achieves the highest combined style‑meaning score among baselines, including GPT‑5.1, and improves meaning preservation, as confirmed by human evaluation.
By Sarubi Thillainathan, Ji-Ung Lee, Michael Sullivan, Alexander Koller
The paper investigates how to balance retaining past experience versus learning from new data when robot dynamics change. It introduces two metrics—change magnitude and age‑staleness AUC—to quantify when older transitions are helpful or harmful. Experiments on locomotion tasks and real‑world perturbations show that the optimal replay strategy depends on the size of the dynamics shift and the evolution of the system over time.
By Everest Yang, Skye Thompson, George D. Konidaris
The paper presents Real‑Time EXPO‑FT, a reinforcement learning framework that fine‑tunes large Vision‑Language‑Action models for real‑time robotic control. It separates slow, expressive action generation from fast, reactive edits, allowing a lightweight policy to adjust actions based on the latest observation. Experiments on the Kinetix benchmark and four dynamic real‑world tasks show that Real‑Time EXPO‑FT achieves superior performance, improving policy success rates from 42% to 97% with only ten minutes of online data and no human intervention.
By Perry Dong, Kuo-Han Hung, Dorsa Sadigh, Chelsea Finn
arXiv:2609.18642v1 Announce Type: new
Abstract: Large language models (LLMs) often suffer from capability stagnation in self-improvement training because fixed difficulty levels fail to adapt to thei...
By Yajie Yu, Mark Lee, Yue Feng
arXiv:2609.17552v1 Announce Type: new
Abstract: Moral-reward RL can make language-model agents more cooperative, but whether that alignment survives adversarial persona pressure is unknown. Such atta...
By Arth Singh
arXiv:2609.17848v1 Announce Type: new
Abstract: Limited controlled evidence exists on how training data, adaptation method, and model scale jointly affect tool-calling performance in language-model a...
By Md Tahmid Rahman Laskar, Xue-Yong Fu, Shashi Bhushan TN
arXiv:2609.18677v1 Announce Type: new
Abstract: Speech language models enable richer spoken interactions between humans and machines than cascaded systems, allowing access to paralinguistic informati...
By Timoth\'ee Weisselberger, Edouard Graves, Alexandre D\'efossez
arXiv:2609.19134v1 Announce Type: new
Abstract: Scientific code repositories encode decades of human knowledge in executable models, methods, and tools. Yet fragmented toolchains, implicit domain con...
By Hejia Geng, Zesen Huang, Haoyang Li, Wenbin Li, Koutian Wu, Zihan Zhou, Yuanbo Pang, Weihao Liu, Zigong Xu, Zhiping Li, Zongzheng Zhang, Chuanfei Dong, Jiankai Sun, Tianzhe Zheng, Fengyu Xie, Yue Ma, Yueheng Shi, Tong Xie, Zonglin Di, Xianrong Liu, Qucheng Gao, Yimin Liu, Jiaming Pan, Sheng Huang, Xiao-Han Ma, Lanqing Yuan, Zhenlin Zhu, Ziang Liu, Ziyang Xu, Junkai Wang, Kangkai Liang, Jiayi Xian, Zehong Zhao, Liuwei Xu, Jingxu Xie, Peijin Zhang, Qiang Gao, Chengyi Xing, Zhe Zhao, Xi Wang, Yaopeng Xing, Xing Meng, Zhenfei Yin, Yingcheng Wu, Ling Yang
arXiv:2609.17800v1 Announce Type: new
Abstract: Vision-language models (VLMs) remain largely unreliable on panoramic dental radiographs and can rely on learned anatomical priors rather than evidence...
By Ahmed Rafid, Fariya Ahmed, Rumman Adib, Mehedi Ahamed, Ajwad Abrar, Tareque Mohmud Chowdhury
arXiv:2602.02377v3 Announce Type: replace
Abstract: While large language models (LLMs) have achieved strong performance on mathematical problems with verifiable answers, many advanced problems are pr...
By Haotong Yang, Zitong Wang, Shijia Kang, Siqi Yang, Wenkai Yu, Xu Niu, Yike Sun, Yi Hu, Zhouchen Lin, Muhan Zhang
arXiv:2607.28077v2 Announce Type: replace
Abstract: Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models, but prompt groups with identica...
By Shuang Liang, Haoyang Zhou, Yifan Gong, Guowei Wang, Xiting Wang
The paper develops a geometric theory of decision boundaries for structured Markov Decision Processes, treating the geometry induced by optimal policies as the key analytical object. It shows that, under structural regularity, this geometry yields the minimal representation needed for policy reconstruction and dictates the statistical and computational complexity of the reconstruction problem. The authors introduce intrinsic notions of boundary and decision complexity, derive information-theoretic measures of decision compression, and provide statistical guarantees for boundary estimation and policy reconstruction from black-box queries, supported by controlled numerical experiments.
By Fredy Pokou (MRE, INOCS)
The paper introduces Composite‑Gradient Learning (CGL), a method that explicitly incorporates a model predictive controller (MPC) into the training of a deep reinforcement learning (DRL) agent by treating their control inputs as a joint action. CGL updates the DRL policy while accounting for the interaction with the MPC, unlike prior approaches that view MPC merely as part of the environment. Experiments on two freeway traffic networks show that CGL performs better than alternative methods when the interaction between DRL and MPC is strong, though overall gains are modest.
By Giray \"On\"ur, Azita Dabiri, Bart De Schutter
Delphos is a multitask reinforcement learning framework that automates discrete choice model specification by treating it as a sequential decision-making problem. It learns transferable specification strategies across multiple transport choice datasets using a DeepSet‑Q architecture, enabling a shared policy to adapt to varying variable sets. Trained on nine datasets, Delphos outperforms single‑task agents, and when applied to unseen Swissmetro and Decisions datasets, it quickly identifies competitive specifications with higher log‑likelihoods than existing methods.
By Gabriel Nova, Stephane Hess, Sander Van Cranenburgh