MOFjson is a new representation for metal‑organic frameworks that separates structural data into a raw atomic layer with explicit bond shift vectors and a derived layer that exposes secondary building units and network topology. The authors benchmarked MOFjson against three baselines across four question categories, achieving superior performance (e.g., 100% vs. 18‑47% for nearest periodic copy detection). Ablation of shift labels reduced accuracy dramatically, demonstrating that models rely on the periodic graph rather than surface format. MOFjson provides language models with a structured, topological view of MOFs, enabling more accurate predictions and facilitating agentic tool workflows and constrained MOF generation.
By Anabel Yong, Matthew Hart, Philippe Schwaller
The paper introduces a conditional diffusion framework that generates synthetic Doppler velocity log (DVL) and inertial sensor signals for autonomous underwater vehicle (AUV) navigation. Using an elucidated diffusion model conditioned on three per‑channel motion descriptors, the authors evaluate three configurations—DVL beams alone, inertial channels alone, and all channels jointly—across thirteen real trajectories, achieving Pearson correlations between 0.72 and 0.997. A BeamsNet velocity estimator trained solely on these generated signals matches the performance of one trained on the full real dataset, indicating that synthetic data can replace real recordings without measurable loss.
By Michal Levin, Itzik Klein
Adaptive Memory Index Evolution for Agents (AMIE) is a lightweight plug‑in layer that evolves per‑memory access state from deployment retrievals without altering stored content or retraining the host memory system. It distills a semantic teacher’s keep/suppress judgments into a retrieval‑conditioned policy and accumulates selective actions into bounded state that reweights subsequent host retrievals. Across five retrieval‑based hosts and three benchmarks, trained AMIE improves downstream F1 or accuracy in all 15 comparisons (0.70–3.37 percentage points) and improves retrieval in all 10 comparisons, demonstrating that persistent memory access can be adapted independently of memory content.
By Yuntong Hu, Tongli Su, Yuanfu Sun, Shuo Ma, Qiaoyu Tan, Liang Zhao
The paper evaluates a spiking version of the MoveNet architecture, called S-MoveNet-Lightning, for autonomous spacecraft pose estimation. Using the RAPTOR dataset, it compares the spiking model to the standard MoveNet-Lightning, finding that the spiking version requires on average 2.2 times fewer multiplication and accumulation operations on FPGA platforms with AMD Deep Learning Processor Units. The study highlights balanced advantages in computational resources and architectural settings for spiking U‑Net‑like architectures in embedded, event‑driven space applications.
By Jonathan Courtois, Christophe Marabotto, Benoit Miramond, Alain Pegatoquet
The survey reviews uncertainty quantification in financial foundation models, noting that uncertainty arises at multiple stages of financial workflows—from evidence and context to model inference and outcome/action. It introduces a source‑aware framework that categorizes six uncertainty sources and a mechanism‑based taxonomy for how methods construct or verify uncertainty representations. The authors map existing finance‑specific studies onto this taxonomy, finding that most research focuses on isolated workflow components and that few studies link evidence reliability to downstream predictions and constrained actions.
By Zijie Zhao, Mingjun Sun, Shenbo Xu
IntNet is a multi-agent reinforcement learning framework designed to improve safety in autonomous driving by integrating vehicle intent transmission and adaptive communication scheduling into a single end-to-end learning system. The model uses an observation prediction module for agents to forecast future policy outputs and share them over a vehicle-to-vehicle network, while a Graph Attention Network processes incoming messages and a scheduler optimizes the communication graph for bandwidth efficiency. Experiments in complex urban environments show that IntNet achieves better coordination, lower collision rates, and reduces communication effort by up to 60% compared to state‑of‑the‑art communicative MARL methods.
By Leandro Parada, Kevin Yu, Panagiotis Angeloudis
Dream2Assist is a model‑based reinforcement learning framework that extends a recurrent state‑space world model with intent inference to support shared vehicle control. The system is trained via fictitious co‑play to anticipate discrete human objectives without overriding agency, and includes a full deployment pipeline to translate the policy into a safe, human‑centered real‑world system. Validation through three human‑subject studies and on‑vehicle deployments in high‑speed racing and urban driving shows that intent conditioning improves torque alignment and reduces collision rates.
By Jonathan DeCastro, Thomas Balch, Emily Sumner, Deepak Gopinath, Laporsha Dees, Patricio Reyes Gomez, Elliot Weiss, Megan Applegate-Kenton, Todd Rowell, John Leonard, Avinash Balachandran, Guy Rosman
The paper introduces Meta-Gradient Informed Multi-Objective Multi-Agent Reinforcement Learning (MIRL), a framework that leverages the structure of multiple objectives to enhance cooperation among agents. By formulating MOMARL at both team and individual levels and using objective-wise counterfactual credit assignment, MIRL allows agents to learn how to prioritize objectives individually while jointly optimizing a system-level objective. Experiments on a multi-objective particle environment demonstrate that MIRL, especially its dynamic variant, improves Pareto-front quality compared to baseline methods.
By Qiyue Xia, J. Michael Herrmann
The paper surveys 86 recent studies on the security of reinforcement learning (RL) in autonomous driving. It systematically categorizes attacks and defenses by threat models and single- versus multi-agent settings, addressing gaps in previous surveys. The authors also evaluate how current attacks and defenses apply to autonomous driving, offering guidance for building robust RL systems.
By Ambra Demontis, Srishti Gupta, Maura Pintor, Luca Demetrio, Kathrin Grosse, Hsiao-Ying Lin, Chengfang Fang, Battista Biggio, Fabio Roli
The paper investigates online learning for multiple strategic agents interacting within a shared linear dynamical system that experiences adversarial disturbances. Each agent aims to minimize its own convex loss while observing the full state, and the study extends the stabilizing-baseline approach from nonstochastic online control to this multi‑agent setting. By employing online gradient descent with memory under state‑only and aggregate‑input feedback, the authors establish sublinear, near‑optimal per‑agent regret bounds and, in the common‑interest case, demonstrate equilibrium‑tracking guarantees via a time‑varying potential game framework.
By Anas Barakat, John Lazarsfeld, Georgios Piliouras, Antonios Varvitsiotis
FlexMARL is a graph-driven rollout‑training co‑designed framework that addresses the infrastructure gap in large‑scale LLM‑based multi‑agent reinforcement learning. It introduces a multi‑agent execution graph (MEG) and a MEG Store for decentralized storage, a Rollout Engine that enables dependency‑parallel sampling with hierarchical load balancing, a Joint Orchestrator that uses an agent‑level asynchronous pipeline to reduce synchronization barriers, and a Training Engine that allocates resources on demand and swaps training states via location‑agnostic APIs. Experiments on a production cluster show up to 7.3× speedup and 5.6× better hardware utilization compared to existing frameworks.
By Zhida Jiang, Zhaolong Xing, Jiawei Lu, Yipei Niu, Qingyuan Sang, Liangxu Zhang, Wenquan Dai, Junhua Shu, Jiaxing Wang, Qiangyu Pei, Qiong Chen, Xinyu Liu, Fangming Liu, Ai Han, Zhen Chen, Ke Zhang
HARTS is a training system that accelerates agentic reinforcement learning with hybrid‑attention models by sharing prefixes of irregular rollout trees. It uses a two‑level planner to create compression‑oriented microbatches, refines their membership and execution order, and accumulates gradients before a single backward pass through shared prefixes. The system supports distributed parallelism, activation recomputation, and preserves RL objectives while achieving 4.77‑10.45× higher update throughput on large hybrid‑attention MoE models.
By Boyuan Meng, Peihua Bao, Hong Liu, Xiaowei Zhu, Chao Wang, Gen Li, Zhenxuan Pan
OmniBrain-0 is a lightweight, unified embodied agent that coordinates manipulation and navigation through a shared high‑level decision model. It integrates task understanding, planning, progress tracking, and feedback‑driven recovery, and is trained via embodied pretraining, domain‑specialized agent distillation, and multi‑teacher on‑policy distillation. The framework also introduces Rubric‑Guided Recursive Self‑Improvement, a data flywheel that refines rubrics and trajectories for continual domain‑expert updates, enabling efficient transfer and consolidation of embodied capabilities across multiple benchmarks and real‑world tasks.
By Xiaomi Embodied Intelligence Team, Yuechen Luo (University of Macau, Tsinghua University), Chong He (University of Macau, Tsinghua University), Fang Li (University of Macau, Tsinghua University), Shaoqing Xu (University of Macau, Tsinghua University), Jingkai Wang (University of Macau, Tsinghua University), Menglun Su (University of Macau, Tsinghua University), Jiaxin Liu (University of Macau, Tsinghua University), Shengyin Jiang (University of Macau, Tsinghua University), Zhiyi Lai (University of Macau, Tsinghua University), Long Chen (University of Macau, Tsinghua University), Longlong Wang (University of Macau, Tsinghua University), Liang Hu (University of Macau, Tsinghua University), Haochen Liu (University of Macau, Tsinghua University), Yuncheng Jiang (University of Macau, Tsinghua University), Hongwei Xie (University of Macau, Tsinghua University), Jieming Cui (University of Macau, Tsinghua University), Hanbing Li (University of Macau, Tsinghua University), Junjie Huang (University of Macau, Tsinghua University), Zhixiang Duan (University of Macau, Tsinghua University), Hexin Zhang (University of Macau, Tsinghua University), Weixiang Liang (University of Macau, Tsinghua University), Ying Gong (University of Macau, Tsinghua University), Yuhan Wang (University of Macau, Tsinghua University), Yong Pan (University of Macau, Tsinghua University), Pengna Li (University of Macau, Tsinghua University), Kangyi Wu (University of Macau, Tsinghua University), Ken Deng (University of Macau, Tsinghua University), Kun Ma (University of Macau, Tsinghua University), Fuxi Wen (University of Macau, Tsinghua University), Zhi-xin Yang (University of Macau, Tsinghua University), Hangjun Ye (University of Macau, Tsinghua University)
The paper introduces a continued‑pretraining corpus called "coupled welfare" that embeds ecological reasoning into large language models. After a single QLoRA run, the corpus was applied to five open models across three architectures, resulting in a dramatic reduction of harmful decisions under high‑pressure scenarios while leaving general performance unchanged. The study also demonstrates that the improvement is content‑specific, as a neutral control corpus did not yield similar gains, and that the models remain robust against adversarial attempts to justify environmentally destructive actions.
By Trent R. Northen, Mingxun Wang
The paper introduces a metacognitive controller that optimizes diffusion language models for agent tasks by deciding when to finalize tokens and how far to generate, thereby reducing computation. The lightweight plug‑in is trained on agent trajectories to balance correctness and generation cost. Experiments on five benchmarks show an average 20.6% reduction in inference FLOPs while keeping task accuracy competitive.
By Mingyu Zhao, Shuhang Lin, Ligong Han, Xiaojing Yu, Zhenting Wang, Yifei Pang, Yang Zhou, Dimitris N. Metaxas
The paper investigates how language‑model agents learn from the harnesses—prompts, tools, and control logic—through which they are deployed. Using KL‑regularized control, it shows that a trained policy’s log odds reflect the harness’s action preferences, and that the agent’s deployed return equals the harness’s best possible return minus a mismatch term caused by the policy’s suboptimal choices. Experiments on MiniWoB++ and Terminal‑Bench Lite demonstrate that changing harnesses can drastically alter performance, and that the ceiling‑minus‑mismatch metric predicts gains better than either component alone.
By Jian Mu, Ruibo Shen, Yao Shu
The paper introduces the Preference Dependency Graph Framework (PDG‑F) for narrative‑driven recommendation, which parses free‑form user requests into a directed acyclic graph of preference dependencies. PDG‑A constructs this graph, enabling directed path retrieval of candidate evidence from the web and topological ranking to produce final recommendations. Experiments across English and Chinese domains show PDG‑F outperforms baselines, highlighting the importance of modeling hierarchical preference relationships.
By Mingzhe Lu, Jian Qin, Ruize Qin, Jinhao Liu, Yunpeng Li, Hang Su
The paper introduces a unified framework called cascade routing that combines routing and cascading strategies for selecting large language models (LLMs) in agentic systems. It derives an optimal cascading strategy, proves the optimality of an existing routing strategy, and demonstrates that integrating both leads to a theoretically optimal approach. Experiments show cascade routing consistently outperforms the individual methods, and the study identifies high‑quality estimators as crucial for effective model selection.
By Jasper Dekoninck, Maximilian Baader, Martin Vechev
arXiv:2610. 37657v1 Announce Type: cross Abstract: We introduce Heads of Security (HoS), a prompt injection defense that enforces instruction-data separation via a small set of learned per-head attention gates.
By Igor Shilov, Xiaoxue Yang, Alexander Panfilov, Evgenii Kortukov, Matthieu Meeus, Peter Romov, Wojciech Samek, Jonas Geiping, Yves-Alexandre de Montjoye
This systematic review examines the use of natural language processing (NLP) throughout the entire battery life cycle, rather than focusing on a single stage or method. It introduces a new technical language processing (TLP) framework tailored for the European Union’s digital battery passport and other battery prediction tasks. The study follows PRISMA guidelines, searches three major databases, and ultimately evaluates 66 relevant papers out of 274, highlighting emerging NLP tasks that aid materials discovery while noting challenges such as the absence of standard benchmarks.
By Tosin Adewumi, Martin Karlsson, Marcus Liwicki, Mikael Sj\"odahl, Lama Alkhaled, Rihab Gargouri, Nudrat Habib