Daily Papers Arch&EAI

Vision-Language-Action (VLA) models combine perception, language, and motor control in a single architecture, yet how they translate multimodal inputs into actions remains poorly understood. We apply activation injection, sparse autoencoders (SAEs), and linear probes to six models spanning 80M--7B parameters across 394,000+ rollout episodes on four benchmarks. The visual pathway dominates action generation across all architectures: injecting baseline activations into null-prompt episodes recovers near-identical behavior, while cross-task injection steers robots toward source-task positions (99.8\% of X-VLA episodes align with the source trajectory), exposing spatially bound motor programs tied to scene coordinates rather than abstract task representations. Language sensitivity depends on task structure, not model design: when visual context uniquely specifies the task, language is ignored; when multiple goals share a scene, language becomes essential (X-VLA \texttt{libero\_goal}: 94\%$\to$10\% under wrong prompts vs.\ \texttt{libero\_object}: 60--100\% regardless). In all three multi-pathway architectures (\pizhalf{}, SmolVLA, GR00T), expert pathways encode motor programs while VLM pathways encode goal semantics ($2\times$ greater behavioral displacement from expert injection), and subspace injection confirms these occupy separable activation subspaces. Per-token SAE processing is essential for action fidelity on most architectures, though mean-pooling improves fidelity on X-VLA. Contrastive identification recovers 82+ manipulation concepts, and causal ablation reveals sensitivity spanning 28--92\% zero-effect rates independent of representation width. We release \textbf{Action Atlas} (https://action-atlas.com) for interactive exploration of VLA representations across all six models.

Vision-Language-Action (VLA) models combine perception, language, and motor control in a single architecture, yet how they translate multimodal inputs into actions remains poorly understood.

There are two major categories of embodied navigation: Vision-Language Navigation (VLN), where agents navigate by following natural language instructions; and Object-Goal Navigation (OGN), where agents navigate to a specified target object. However, existing work primarily evaluates model performance under nominal conditions, overlooking the potential corruptions that arise in real-world settings. To address this gap, we present NavTrust, a unified benchmark that systematically corrupts input modalities, including RGB, depth, and instructions, in realistic scenarios and evaluates their impact on navigation performance. To our best knowledge, NavTrust is the first benchmark that exposes embodied navigation agents to diverse RGB-Depth corruptions and instruction variations in a unified framework. Our extensive evaluation of seven state-of-the-art approaches reveals substantial performance degradation under realistic corruptions, which highlights critical robustness gaps and provides a roadmap toward more trustworthy embodied navigation systems. Furthermore, we systematically evaluate four distinct mitigation strategies to enhance robustness against RGB-Depth and instructions corruptions. Our base models include Uni-NaVid and ETPNav. We deployed them on a real mobile robot and observed improved robustness to corruptions. The project website is: https://navtrust.github.io.

There are two major categories of embodied navigation: Vision-Language Navigation (VLN), where agents navigate by following natural language instructions; and Object-Goal Navigation (OGN), where agents navigate to a specified target object.

We present F2LLM-v2, a new family of general-purpose, multilingual embedding models in 8 distinct sizes ranging from 80M to 14B. Trained on a newly curated composite of 60 million publicly available high-quality data samples, F2LLM-v2 supports more than 200 languages, with a particular emphasis on previously underserved mid- and low-resource languages. By integrating a two-stage LLM-based embedding training pipeline with matryoshka learning, model pruning, and knowledge distillation techniques, we present models that are far more efficient than previous LLM-based embedding models while retaining competitive performances. Extensive evaluations confirm that F2LLM-v2-14B ranks first on 11 MTEB benchmarks, while the smaller models in the family also set a new state of the art for resource-constrained applications. To facilitate open-source embedding model research, we release all models, data, code, and intermediate checkpoints.

We present F2LLM-v2, a new family of general-purpose, multilingual embedding models in 8 distinct sizes ranging from 80M to 14B.

Contact-rich manipulation tasks, such as wiping and assembly, require accurate perception of contact forces, friction changes, and state transitions that cannot be reliably inferred from vision alone. Despite growing interest in visuo-tactile manipulation, progress is constrained by two persistent limitations: existing datasets are small in scale and narrow in task coverage, and current methods treat tactile signals as passive observations rather than using them to model contact dynamics or enable closed-loop control explicitly. In this paper, we present \textbf{OmniViTac}, a large-scale visuo-tactile-action dataset comprising $21{,}000+$ trajectories across $86$ tasks and $100+$ objects, organized into six physics-grounded interaction patterns. Building on this dataset, we propose \textbf{OmniVTA}, a world-model-based visuo-tactile manipulation framework that integrates four tightly coupled modules: a self-supervised tactile encoder, a two-stream visuo-tactile world model for predicting short-horizon contact evolution, a contact-aware fusion policy for action generation, and a 60Hz reflexive controller that corrects deviations between predicted and observed tactile signals in a closed loop. Real-robot experiments across all six interaction categories show that OmniVTA outperforms existing methods and generalizes well to unseen objects and geometric configurations, confirming the value of combining predictive contact modeling with high-frequency tactile feedback for contact-rich manipulation. All data, models, and code will be made publicly available on the project website at https://mrsecant.github.io/OmniVTA.

Contact-rich manipulation tasks, such as wiping and assembly, require accurate perception of contact forces, friction changes, and state transitions that cannot be reliably inferred from vision alone.

Real-time execution is crucial for deploying Vision-Language-Action (VLA) models in the physical world. Existing asynchronous inference methods primarily optimize trajectory smoothness, but neglect the critical latency in reacting to environmental changes. By rethinking the notion of reaction in action chunking policies, this paper presents a systematic analysis of the factors governing reaction time. We show that reaction time follows a uniform distribution determined jointly by the Time to First Action (TTFA) and the execution horizon. Moreover, we reveal that the standard practice of applying a constant schedule in flow-based VLAs can be inefficient and forces the system to complete all sampling steps before any movement can start, forming the bottleneck in reaction latency. To overcome this issue, we propose Fast Action Sampling for ImmediaTE Reaction (FASTER). By introducing a Horizon-Aware Schedule, FASTER adaptively prioritizes near-term actions during flow sampling, compressing the denoising of the immediate reaction by tenfold (e.g., in $π_{0.5}$ and X-VLA) into a single step, while preserving the quality of long-horizon trajectory. Coupled with a streaming client-server pipeline, FASTER substantially reduces the effective reaction latency on real robots, especially when deployed on consumer-grade GPUs. Real-world experiments, including a highly dynamic table tennis task, prove that FASTER unlocks unprecedented real-time responsiveness for generalist policies, enabling rapid generation of accurate and smooth trajectories.

Real-time execution is crucial for deploying Vision-Language-Action (VLA) models in the physical world.

Vision-Language-Action (VLA) models have emerged as a promising approach for general-purpose robot manipulation. However, their generalization is inconsistent: while these models can perform impressively in some settings, fine-tuned variants often fail on novel objects, scenes, and instructions. We apply mechanistic interpretability techniques to better understand the inner workings of VLA models. To probe internal representations, we train Sparse Autoencoders (SAEs) on hidden layer activations of the VLA. SAEs learn a sparse dictionary whose features act as a compact, interpretable basis for the model's computation. We find that the large majority of extracted SAE features correspond to memorized sequences from specific training demonstrations. However, some features correspond to interpretable, general, and steerable motion primitives and semantic properties, offering a promising glimpse toward VLA generalizability. We propose a metric to categorize features according to whether they represent generalizable transferable primitives or episode-specific memorization. We validate these findings through steering experiments on the LIBERO benchmark. We show that individual SAE features causally influence robot behavior. Steering general features induces behaviors consistent with their semantic meaning and can be applied across tasks and scenes. This work provides the first mechanistic evidence that VLAs can learn generalizable features across tasks and scenes. We observe that supervised fine-tuning on small robotics datasets disproportionately amplifies memorization. In contrast, training on larger, more diverse datasets (e.g., DROID) or using knowledge insulation promotes more general features. We provide an open-source codebase and user-friendly interface for activation collection, SAE training, and feature steering. Our project page is located at http://drvla.github.io

Vision-Language-Action (VLA) models have emerged as a promising approach for general-purpose robot manipulation.

Despite the computational efficiency of MoE models, the excessive memory footprint and I/O overhead inherent in multi-expert architectures pose formidable challenges for real-time inference on resource-constrained edge platforms. While existing static methods struggle with a rigid latency-accuracy trade-off, we observe that expert importance is highly skewed and depth-dependent. Motivated by these insights, we propose DyMoE, a dynamic mixed-precision quantization framework designed for high-performance edge inference. Leveraging insights into expert importance skewness and depth-dependent sensitivity, DyMoE introduces: (1) importance-aware prioritization to dynamically quantize experts at runtime; (2) depth-adaptive scheduling to preserve semantic integrity in critical layers; and (3) look-ahead prefetching to overlap I/O stalls. Experimental results on commercial edge hardware show that DyMoE reduces Time-to-First-Token (TTFT) by 3.44x-22.7x and up to a 14.58x speedup in Time-Per-Output-Token (TPOT) compared to state-of-the-art offloading baselines, enabling real-time, accuracy-preserving MoE inference on resource-constrained edge devices.

Despite the computational efficiency of MoE models, the excessive memory footprint and I/O overhead inherent in multi-expert architectures pose formidable challenges for real-time inference on resource-constrained edge platforms.

This paper proposes a fully decentralized model predictive control (MPC) framework with control barrier function (CBF) constraints for safety-critical trajectory planning in multi-robot legged systems. The incorporation of CBF constraints introduces explicit inter-agent coupling, which prevents direct decomposition of the resulting optimal control problems. To address this challenge, we reformulate the centralized safety-critical MPC problem using a structured distributed optimization framework based on the alternating direction method of multipliers (ADMM). By introducing a novel node-edge splitting formulation with consensus constraints, the proposed approach decomposes the global problem into independent node-local and edge-local quadratic programs that can be solved in parallel using only neighbor-to-neighbor communication. This enables fully decentralized trajectory optimization with symmetric computational load across agents while preserving safety and dynamic feasibility. The proposed framework is integrated into a hierarchical locomotion control architecture for quadrupedal robots, combining high-level distributed trajectory planning, mid-level nonlinear MPC enforcing single rigid body dynamics, and low-level whole-body control enforcing full-order robot dynamics. The effectiveness of the proposed approach is demonstrated through hardware experiments on two Unitree Go2 quadrupedal robots and numerical simulations involving up to four robots navigating uncertain environments with rough terrain and external disturbances. The results show that the proposed distributed formulation achieves performance comparable to centralized MPC while reducing the average per-cycle planning time by up to 51% in the four-agent case, enabling efficient real-time decentralized implementation.

This paper proposes a fully decentralized model predictive control (MPC) framework with control barrier function (CBF) constraints for safety-critical trajectory planning in multi-robot legged systems.

Large Reasoning Models (LRMs) have shown remarkable reasoning capabilities, yet they often suffer from overthinking, expending redundant computational steps on simple problems, or underthinking, failing to explore sufficient reasoning paths despite inherent capabilities. These issues lead to inefficiencies and potential inaccuracies, limiting practical deployment in resource-constrained settings. Existing methods to mitigate overthinking, such as suppressing reflective keywords or adjusting reasoning length, may inadvertently induce underthinking, compromising accuracy. Therefore, we propose ReBalance, a training-free framework that achieves efficient reasoning with balanced thinking. ReBalance leverages confidence as a continuous indicator of reasoning dynamics, identifying overthinking through high confidence variance and underthinking via consistent overconfidence. By aggregating hidden states from a small-scale dataset into reasoning mode prototypes, we compute a steering vector to guide LRMs' reasoning trajectories. A dynamic control function modulates this vector's strength and direction based on real-time confidence, pruning redundancy during overthinking, and promoting exploration during underthinking. Extensive experiments conducted on four models ranging from 0.5B to 32B, and across nine benchmarks in math reasoning, general question answering, and coding tasks demonstrate that ReBalance effectively reduces output redundancy while improving accuracy, offering a general, training-free, and plug-and-play strategy for efficient and robust LRM deployment. Project page and code are available at https://rebalance-ai.github.io .

Large Reasoning Models (LRMs) have shown remarkable reasoning capabilities, yet they often suffer from overthinking, expending redundant computational steps on simple problems, or underthinking, failing to explore sufficient reasoning paths despite inherent capabilities.

Vision-Language-Action (VLA) models have recently enabled embodied agents to perform increasingly complex tasks by jointly reasoning over visual, linguistic, and motor modalities. However, we find that the prevailing notion of ``efficiency'' in current VLA research, characterized by parameters, FLOPs, or token decoding throughput, does not reflect actual performance on robotic platforms. In real-world execution, efficiency is determined by system-level embodied behaviors such as task completion time, trajectory smoothness, cumulative joint rotation, and motion energy. Through controlled studies across model compression, token sparsification, and action sequence compression, we make several observations that challenge common assumptions. (1) Methods that reduce computation under conventional metrics often increase end-to-end execution cost or degrade motion quality, despite maintaining task success rates. (2) System-level embodied efficiency metrics reveal performance differences in the learned action policies that remain hidden under conventional evaluations. (3) Common adaptation methods such as in-context prompting or supervised fine-tuning show only mild and metric-specific improvements in embodied efficiency. While these methods can reduce targeted embodied-efficiency metrics such as jerk or action rate, the resulting gains may come with trade-offs in other metrics, such as longer completion time. Taken together, our results suggest that conventional inference efficiency metrics can overlook important aspects of embodied execution. Incorporating embodied efficiency provides a more complete view of policy behavior and practical performance, enabling fairer and more comprehensive comparisons of VLA models.

Vision-Language-Action (VLA) models have recently enabled embodied agents to perform increasingly complex tasks by jointly reasoning over visual, linguistic, and motor modalities.

Transformers have revolutionized AI in natural language processing and computer vision, but their large computation and memory demands pose major challenges for hardware acceleration. In practice, end-to-end throughput is often limited by paged data movement and interconnect bandwidth rather than raw MAC count. This work proposes a unified system-accelerator co-design approach for transformer inference that jointly optimizes a matrix accelerator and its system integration through paged streaming dataflows and explicit overlap of compute and transfer. On the hardware side, we introduce MatrixFlow, a loosely coupled 16x16 systolic-array accelerator with a page-aligned block matrix multiplication method using 4 KB tiles, a small on-chip buffer of about 20 KB, and a pipelined schedule of DMA, compute, and DMA-out to utilize interconnect bandwidth efficiently. On the system side, we develop Gem5-AcceSys, an extension of the gem5 full-system simulator that explores standard interconnects such as PCIe and configurable memory hierarchies including Direct Memory, Direct Cache, and Device Memory modes with SMMU/TLB effects. We evaluate the co-design using gem5 simulations on representative transformer models including BERT and ViT across multiple data types and system setups. Results show up to 22x end-to-end speedup over a CPU-only baseline and 5x to 8x gains over state-of-the-art loosely and tightly coupled accelerators. We further show that a standard PCIe-based host-memory design can achieve about 80 percent of the performance of on-device HBM. Overall, paged streaming and pipeline overlap, rather than large local SRAMs, are the most effective levers for efficient transformer inference under realistic system constraints.

Transformers have revolutionized AI in natural language processing and computer vision, but their large computation and memory demands pose major challenges for hardware acceleration.

This work introduces two lightweight model predictive control (MPC) approaches for attitude tracking with reaction wheels during spacecraft rendezvous synchronization. Both approaches are based on a novel attitude deviation formulation, which enables the use of inherently linear constraints on angular velocity. We develop a single-loop and a dual-loop MPC; the latter embeds a stabilizing feedback controller within the inner loop, yielding a linear time-invariant system. Both controllers are implemented with CasADi - including automatic code generation - evaluated across various solvers, and validated within the Basilisk astrodynamics simulation framework. The experimental results demonstrate improved tracking accuracy alongside reductions in computational effort and memory consumption. Finally, embedded delivery to an ARM Cortex-M7 - representative of commercial off-the-shelf devices used in New Space platforms - confirms the real-time feasibility of these approaches and highlights their suitability for onboard attitude control in resource-constrained spacecraft rendezvous missions.

This work introduces two lightweight model predictive control (MPC) approaches for attitude tracking with reaction wheels during spacecraft rendezvous synchronization.

Spatial reasoning is foundational for Vision-Language Models (VLMs), particularly when deployed as Vision-Language-Action (VLA) agents in physical environments. However, existing benchmarks predominantly focus on elementary, single-hop relations, neglecting the multi-hop compositional reasoning and precise visual grounding essential for real-world scenarios. To address this, we introduce MultihopSpatial, offering three key contributions: (1) A comprehensive benchmark designed for multi-hop and compositional spatial reasoning, featuring 1- to 3-hop complex queries across diverse spatial perspectives. (2) Acc@50IoU, a complementary metric that simultaneously evaluates reasoning and visual grounding by requiring both answer selection and precise bounding box prediction - capabilities vital for robust VLA deployment. (3) MultihopSpatial-Train, a dedicated large-scale training corpus to foster spatial intelligence. Extensive evaluation of 37 state-of-the-art VLMs yields eight key insights, revealing that compositional spatial reasoning remains a formidable challenge. Finally, we demonstrate that reinforcement learning post-training on our corpus enhances both intrinsic VLM spatial reasoning and downstream embodied manipulation performance.

Spatial reasoning is foundational for Vision-Language Models (VLMs), particularly when deployed as Vision-Language-Action (VLA) agents in physical environments.

In real-world Federated Learning (FL) deployments, data distributions on devices that participate in training evolve over time. This leads to asynchronous data drift, where different devices shift at different times and toward different distributions. Mitigating such drift is challenging: frequent retraining incurs high computational cost on resource-constrained devices, while infrequent retraining degrades performance on drifting devices. We propose DriftGuard, a federated continual learning framework that efficiently adapts to asynchronous data drift. DriftGuard adopts a Mixture-of-Experts (MoE) inspired architecture that separates shared parameters, which capture globally transferable knowledge, from local parameters that adapt to group-specific distributions. This design enables two complementary retraining strategies: (i) global retraining, which updates the shared parameters when system-wide drift is identified, and (ii) group retraining, which selectively updates local parameters for clusters of devices identified via MoE gating patterns, without sharing raw data. Experiments across multiple datasets and models show that DriftGuard matches or exceeds state-of-the-art accuracy while reducing total retraining cost by up to 83%. As a result, it achieves the highest accuracy per unit retraining cost, improving over the strongest baseline by up to 2.3x. DriftGuard is available for download from https://github.com/blessonvar/DriftGuard.

In real-world Federated Learning (FL) deployments, data distributions on devices that participate in training evolve over time.

Autonomous aerial vehicles (AAVs) empower sixth-generation (6G) Internet-of-Things (IoT) networks through mobility-driven data collection. However, conventional reward-driven reinforcement learning for AAV trajectory planning suffers from severe credit assignment issues and training instability, because sparse scalar rewards fail to capture the long-term and nonlinear effects of sequential movements. To address these challenges, this paper proposes Learn for Variation (L4V), a gradient-informed trajectory learning framework that replaces high-variance scalar reward signals with dense and analytically grounded policy gradients. Particularly, the coupled evolution of AAV kinematics, distance-dependent channel gains, and per-user data-collection progress is first unrolled into an end-to-end differentiable computational graph. Backpropagation through time then serves as a discrete adjoint solver, which propagates exact sensitivities from the cumulative mission objective to every control action and policy parameter. These structured gradients are used to train a deterministic neural policy with temporal smoothness regularization and gradient clipping. Extensive simulations demonstrate that L4V consistently outperforms representative baselines, including a genetic algorithm, DQN, A2C, and DDPG, in mission completion time, average transmission rate, and training cost

Autonomous aerial vehicles (AAVs) empower sixth-generation (6G) Internet-of-Things (IoT) networks through mobility-driven data collection.

Recent Vision-Language-Action models show potential to generalize across embodiments but struggle to quickly align with a new robot's action space when high-quality demonstrations are scarce, especially for bipedal humanoids. We present TrajBooster, a cross-embodiment framework that leverages abundant wheeled-humanoid data to boost bipedal VLA. Our key idea is to use end-effector trajectories as a morphology-agnostic interface. TrajBooster (i) extracts 6D dual-arm end-effector trajectories from real-world wheeled humanoids, (ii) retargets them in simulation to Unitree G1 with a whole-body controller trained via a heuristic-enhanced harmonized online DAgger to lift low-dimensional trajectory references into feasible high-dimensional whole-body actions, and (iii) forms heterogeneous triplets that couple source vision/language with target humanoid-compatible actions to post-pre-train a VLA, followed by only 10 minutes of teleoperation data collection on the target humanoid domain. Deployed on Unitree G1, our policy achieves beyond-tabletop household tasks, enabling squatting, cross-height manipulation, and coordinated whole-body motion with markedly improved robustness and generalization. Results show that TrajBooster allows existing wheeled-humanoid data to efficiently strengthen bipedal humanoid VLA performance, reducing reliance on costly same-embodiment data while enhancing action space understanding and zero-shot skill transfer capabilities. For more details, For more details, please refer to our \href{https://jiachengliu3.github.io/TrajBooster/}.

Recent Vision-Language-Action models show potential to generalize across embodiments but struggle to quickly align with a new robot's action space when high-quality demonstrations are scarce, especially for bipedal humanoids.

Classical methods in robot motion planning, such as sampling-based and optimization-based methods, often struggle with scalability towards higher-dimensional state spaces and complex environments. Diffusion models, known for their capability to learn complex, high-dimensional and multi-modal data distributions, provide a promising alternative when applied to motion planning problems and have already shown interesting results. However, most of the current approaches train their model for a single environment, limiting their generalization to environments not seen during training. The techniques that do train a model for multiple environments rely on a specific camera to provide the model with the necessary environmental information and therefore always require that sensor. To effectively adapt to diverse scenarios without the need for retraining, this research proposes Context-Aware Motion Planning Diffusion (CAMPD). CAMPD leverages a classifier-free denoising probabilistic diffusion model, conditioned on sensor-agnostic contextual information. An attention mechanism, integrated in the well-known U-Net architecture, conditions the model on an arbitrary number of contextual parameters. CAMPD is evaluated on a 7-DoF robot manipulator and benchmarked against state-of-the-art approaches on real-world tasks, showing its ability to generalize to unseen environments and generate high-quality, multi-modal trajectories, at a fraction of the time required by existing methods.

Classical methods in robot motion planning, such as sampling-based and optimization-based methods, often struggle with scalability towards higher-dimensional state spaces and complex environments.

Medical ultrasound image segmentation faces significant challenges due to limited labeled data and characteristic imaging artifacts including speckle noise and low-contrast boundaries. While semi-supervised learning (SSL) approaches have emerged to address data scarcity, existing methods suffer from suboptimal unlabeled data utilization and lack robust feature representation mechanisms. In this paper, we propose Switch, a novel SSL framework with two key innovations: (1) Multiscale Switch (MSS) strategy that employs hierarchical patch mixing to achieve uniform spatial coverage; (2) Frequency Domain Switch (FDS) with contrastive learning that performs amplitude switching in Fourier space for robust feature representations. Our framework integrates these components within a teacher-student architecture to effectively leverage both labeled and unlabeled data. Comprehensive evaluation across six diverse ultrasound datasets (lymph nodes, breast lesions, thyroid nodules, and prostate) demonstrates consistent superiority over state-of-the-art methods. At 5\% labeling ratio, Switch achieves remarkable improvements: 80.04\% Dice on LN-INT, 85.52\% Dice on DDTI, and 83.48\% Dice on Prostate datasets, with our semi-supervised approach even exceeding fully supervised baselines. The method maintains parameter efficiency (1.8M parameters) while delivering superior performance, validating its effectiveness for resource-constrained medical imaging applications. The source code is publicly available at https://github.com/jinggqu/Switch

Medical ultrasound image segmentation faces significant challenges due to limited labeled data and characteristic imaging artifacts including speckle noise and low-contrast boundaries.

Direct handset-to-satellite (DHTS) communication is emerging as a core capability of 6G non-terrestrial networks, enabling standard devices to directly access low Earth orbit (LEO) satellites. While LEO provides the physical access layer for DHTS, large-scale device connectivity introduces challenges in mobility management, interference control, spectrum efficiency, and constellation-wide coordination. Relay-only LEO architectures are insufficient to manage massive handset access under dynamic traffic and energy constraints. This article introduces a hierarchical architecture in which direct handset-to-LEO access is supported by multi-orbit space-based data centers (SBDCs) spanning LEO, medium Earth orbit (MEO), and geostationary Earth orbit (GEO). In this framework, LEO satellites handle radio access and real-time inference, while higher orbital layers provide regional aggregation, global orchestration, and compute-aware routing. By embedding distributed in-orbit computing, energy-aware scheduling, and AI-driven hierarchical control, the constellation evolves from a passive relay network into an intelligent multi-layer system capable of supporting large-scale DHTS services. We discuss key enabling technologies, envisioned multi-orbit integrated Earth-space compute architecture, and open research challenges in integrating multi-orbit computing, highlighting pathways toward scalable and resilient 6G DHTS networks.

Direct handset-to-satellite (DHTS) communication is emerging as a core capability of 6G non-terrestrial networks, enabling standard devices to directly access low Earth orbit (LEO) satellites.

The increasing complexity of neural networks poses significant challenges for democratizing FL on resource?constrained client devices. Parallel split learning (PSL) has emerged as a promising solution by offloading substantial computing workload to a server via model partitioning, shrinking client-side computing load, and eliminating the client-side model aggregation for reduced communication and deployment costs. Since PSL is aggregation-free, it suffers from severe training divergence stemming from gradient directional inconsistency across clients. To address this challenge, we propose GAPSL, a gradient-aligned PSL framework that comprises two key components: leader gradient identification (LGI) and gradient direction alignment (GDA). LGI dynamically selects a set of directionally consistent client gradients to construct a leader gradient that captures the global convergence trend. GDA employs a direction-aware regularization to align each client's gradient with the leader gradient, thereby mitigating inter-device gradient directional inconsistency and enhancing model convergence. We evaluate GAPSL on a prototype computing testbed. Extensive experiments demonstrate that GAPSL consistently outperforms state-of-the-art benchmarks in training accuracy and latency.

The increasing complexity of neural networks poses significant challenges for democratizing FL on resource?constrained client devices.

Despite strong generalization capabilities, Vision-Language-Action (VLA) models remain constrained by the high cost of expert demonstrations and insufficient real-world interaction. While online reinforcement learning (RL) has shown promise in improving general foundation models, applying RL to VLA manipulation in real-world settings is still hindered by low exploration efficiency and a restricted exploration space. Through systematic real-world experiments, we observe that the effective exploration space of online RL is closely tied to the data distribution of supervised fine-tuning (SFT). Motivated by this observation, we propose TwinRL, a digital twin-real-world collaborative RL framework designed to scale and guide exploration for VLA models. First, a high-fidelity digital twin is efficiently reconstructed from smartphone-captured scenes, enabling realistic bidirectional transfer between real and simulated environments. During the SFT warm-up stage, we introduce an exploration space expansion strategy using digital twins to broaden the support of the data trajectory distribution. Building on this enhanced initialization, we propose a sim-to-real guided exploration strategy to further accelerate online RL. Specifically, TwinRL performs efficient and parallel online RL in the digital twin prior to deployment, effectively bridging the gap between offline and online training stages. Subsequently, we exploit efficient digital twin sampling to identify failure-prone yet informative configurations, which are used to guide targeted human-in-the-loop rollouts on the real robot. In our experiments, TwinRL approaches 100% success in both in-distribution regions covered by real-world demonstrations and out-of-distribution regions, delivering at least a 30% speedup over prior real-world RL methods and requiring only about 20 minutes on average across four tasks.

Despite strong generalization capabilities, Vision-Language-Action (VLA) models remain constrained by the high cost of expert demonstrations and insufficient real-world interaction.

The strong performance of large vision-language models (VLMs) trained with reinforcement learning (RL) has motivated similar approaches for fine-tuning vision-language-action (VLA) models in robotics. Many recent works fine-tune VLAs directly in the real world to avoid addressing the sim-to-real gap. While real-world RL circumvents sim-to-real issues, it inherently limits the generality of the resulting VLA, as scaling scene and object diversity in the physical world is prohibitively difficult. This leads to the paradoxical outcome of transforming a broadly pretrained model into an overfitted, scene-specific policy. Training in simulation can instead provide access to diverse scenes, but designing those scenes is also costly. In this work, we show that VLAs can be RL fine-tuned without sacrificing generality and with reduced labor by leveraging 3D world generative models. Using these models together with a language-driven scene designer, we generate hundreds of diverse interactive scenes containing unique objects and backgrounds, enabling scalable and highly parallel policy learning. Starting from a pretrained imitation baseline, our approach increases simulation success from 9.7% to 79.8% while achieving a 1.25$\times$ speedup in task completion time. We further demonstrate successful sim-to-real transfer enabled by the quality of the generated digital twins together with domain randomization, improving real-world success from 21.7% to 75% and achieving a 1.13$\times$ speedup. Finally, we further highlight the benefits of leveraging the effectively unlimited data from 3D world generative models through an ablation study showing that increasing scene diversity directly improves zero-shot generalization.

The strong performance of large vision-language models (VLMs) trained with reinforcement learning (RL) has motivated similar approaches for fine-tuning vision-language-action (VLA) models in robotics.

Non-prehensile (NP) manipulation, in which robots alter object states without forming stable grasps (for example, pushing, poking, or sliding), significantly broadens robotic manipulation capabilities when grasping is infeasible or insufficient. However, enabling a unified framework that generalizes across different tasks, objects, and environments while seamlessly integrating non-prehensile and prehensile (P) actions remains challenging: robots must determine when to invoke NP skills, select the appropriate primitive for each context, and compose P and NP strategies into robust, multi-step plans. We introduce ApaptPNP, a vision-language model (VLM)-empowered task and motion planning framework that systematically selects and combines P and NP skills to accomplish diverse manipulation objectives. Our approach leverages a VLM to interpret visual scene observations and textual task descriptions, generating a high-level plan skeleton that prescribes the sequence and coordination of P and NP actions. A digital-twin based object-centric intermediate layer predicts desired object poses, enabling proactive mental rehearsal of manipulation sequences. Finally, a control module synthesizes low-level robot commands, with continuous execution feedback enabling online task plan refinement and adaptive replanning through the VLM. We evaluate ApaptPNP across representative P&NP hybrid manipulation tasks in both simulation and real-world environments. These results underscore the potential of hybrid P&NP manipulation as a crucial step toward general-purpose, human-level robotic manipulation capabilities. Project Website: https://adaptpnp.github.io/

Non-prehensile (NP) manipulation, in which robots alter object states without forming stable grasps (for example, pushing, poking, or sliding), significantly broadens robotic manipulation capabilities when grasping is infeasible or insufficient.

We propose U-Arm, a low-cost and rapidly adaptable leader-follower teleoperation framework designed to interface with most of commercially available robotic arms. Our system supports teleoperation through three structurally distinct 3D-printed leader arms that share consistent control logic, enabling seamless compatibility with diverse commercial robot configurations. Compared with previous open-source leader-follower interfaces, we further optimized both the mechanical design and servo selection, achieving a bill of materials (BOM) cost of only \$50.5 for the 6-DoF leader arm and \$56.8 for the 7-DoF version. To enhance usability, we mitigate the common challenge in controlling redundant degrees of freedom by %engineering methods mechanical and control optimizations. Experimental results demonstrate that U-Arm achieves 39\% higher data collection efficiency and comparable task success rates across multiple manipulation scenarios compared with Joycon, another low-cost teleoperation interface. We have open-sourced all CAD models of three configs and also provided simulation support for validating teleoperation workflows. We also open-sourced real-world manipulation data collected with U-Arm. The project website is https://github.com/MINT-SJTU/LeRobot-Anything-U-Arm.

We propose U-Arm, a low-cost and rapidly adaptable leader-follower teleoperation framework designed to interface with most of commercially available robotic arms.

Memory-augmented robotic policies are essential in handling memory-dependent tasks. However, existing approaches typically rely on simple observation window extensions, struggling to simultaneously achieve precise task state tracking and robust long-horizon retention. To overcome these challenges, inspired by the Atkinson-Shiffrin memory model, we propose MemoAct, a hierarchical memory-based policy that leverages distinct memory tiers to tackle specific bottlenecks. Specifically, lossless short-term memory ensures precise task state tracking, while compressed long-term memory enables robust long-horizon retention. To enrich the evaluation landscape, we construct MemoryRTBench based on RoboTwin 2.0, specifically tailored to assess policy capabilities in task state tracking and long-horizon retention. Extensive experiments across simulated and real-world scenarios demonstrate that MemoAct achieves superior performance compared to both existing Markovian baselines and history-aware policies. The project page is \href{https://tlf-tlf.github.io/MemoActPage/}{available}.

Memory-augmented robotic policies are essential in handling memory-dependent tasks.

Reinforcement learning (RL) for large-scale Vision-Language-Action (VLA) models faces significant challenges in computational efficiency and data acquisition. We propose AcceRL, a fully asynchronous and decoupled RL framework designed to eliminate synchronization barriers by physically isolating training, inference, and rollouts. Crucially, AcceRL is the first to integrate a plug-and-play, trainable world model into a distributed asynchronous RL pipeline to generate virtual experiences. Experiments on the LIBERO benchmark demonstrate that AcceRL achieves state-of-the-art (SOTA) performance. Systematically, it exhibits super-linear scaling in throughput and highly efficient hardware utilization. Algorithmically, the world-model-augmented variant delivers unprecedented sample efficiency and robust training stability in complex control tasks.

Reinforcement learning (RL) for large-scale Vision-Language-Action (VLA) models faces significant challenges in computational efficiency and data acquisition.

As robots increasingly integrate into everyday environments, ensuring their safe navigation around humans becomes imperative. Efficient and safe motion planning requires robots to account for human behavior, particularly in constrained spaces such as grocery stores or care homes, where interactions with multiple individuals are common. Prior research has employed Bayesian frameworks to model human rationality based on navigational intent, enabling the prediction of probabilistic trajectories for planning purposes. In this work, we present a simple yet novel approach for confidence-aware prediction that treats future predictions as particles. This framework is highly parallelized and accelerated on an graphics processing unit (GPU). As a result, this enables longer-term predictions at a frequency of 125 Hz and can be easily extended for multi-human predictions. Compared to existing methods, our implementation supports finer prediction time steps, yielding more granular trajectory forecasts. This enhanced resolution allows motion planners to respond effectively to subtle changes in human behavior. We validate our approach through real-world experiments, demonstrating a robot safely navigating among multiple humans with diverse navigational goals. Our results highlight the methods potential for robust and efficient human-robot coexistence in dynamic environments.

As robots increasingly integrate into everyday environments, ensuring their safe navigation around humans becomes imperative.

Vision-and-Language Navigation (VLN) is shifting from rigid, step-by-step instruction following toward open-vocabulary, goal-oriented autonomy. Achieving this transition without exhaustive routing prompts requires agents to leverage structural priors. While prior work often assumes computationally heavy 2D/3D metric maps, we instead exploit a lightweight, text-based osmAG (OpenStreetMap Area Graph), a floorplan-level topological representation that is easy to obtain and maintain. However, global planning over a prior map alone is brittle in real-world deployments, where local connectivity can change (e.g., closed doors or crowded passages), leading to execution-time failures. To address this gap, we propose a hierarchical navigation framework HaltNav that couples the robust global planning of osmAG with the local exploration and instruction-grounding capability of VLN. Our approach features an MLLM-based brain module, which is capable of high-level task grounding and obstruction awareness. Conditioned on osmAG, the brain converts the global route into a sequence of localized execution snippets, providing the VLN executor with prior-grounded, goal-centric sub-instructions. Meanwhile, it detects local anomalies via a mechanism we term Reactive Visual Halting (RVH), which interrupts the local control loop, updates osmAG by invalidating the corresponding topology, and triggers replanning to orchestrate a viable detour. To train this halting capability efficiently, we introduce a data synthesis pipeline that leverages generative models to inject realistic obstacles into otherwise navigable scenes, substantially enriching hard negative samples. Extensive experiments demonstrate that our hierarchical framework outperforms several baseline methods without tedious language instructions, and significantly improves robustness for long-horizon vision-language navigation under environmental changes.

Vision-and-Language Navigation (VLN) is shifting from rigid, step-by-step instruction following toward open-vocabulary, goal-oriented autonomy.

Stable and reliable grasp is critical to robotic manipulations especially for fragile and glazed objects, where the grasp force requires precise control as too large force possibly damages the objects while small force leads to slip and fall-off. Although it is assumed the objects to manipulate is grasped firmly in advance, slip detection and timely prevention are necessary for a robot in unstructured and universal environments. In this work, we addressed this issue by utilizing multimodal tactile feedback from a five-fingered bio-inspired hand. Motivated by human hands, the tactile sensing elements were distributed and embedded into the soft skin of robotic hand, forming 24 tactile channels in total. Different from the threshold method that was widely employed in most existing works, we converted the slip detection problem to contact status recognition in combination with binning technique first and then detected the slip onset time according to the recognition results. After the 24-channel tactile signals passed through discrete wavelet transform, 17 features were extracted from different time and frequency bands. With the optimal 120 features employed for status recognition, the test accuracy reached 96.39% across three different sliding speeds and six kinds of materials. When applied to four new unseen materials, a high accuracy of 91.95% was still achieved, which further validated the generalization of our proposed method. Finally, the performance of slip detection is verified based on the trained model of contact status recognition.

Stable and reliable grasp is critical to robotic manipulations especially for fragile and glazed objects, where the grasp force requires precise control as too large force possibly damages the objects while small force leads to slip and fall-off.

Robotic manipulation requires explicit or implicit knowledge of the robot's joint positions. Precise proprioception is standard in high-quality industrial robots but is often unavailable in inexpensive robots operating in unstructured environments. In this paper, we ask: to what extent can a fast, single-pass regression architecture perform visual proprioception from a single external camera image, available even in the simplest manipulation settings? We explore several latent representations, including CNNs, VAEs, ViTs, and bags of uncalibrated fiducial markers, using fine-tuning techniques adapted to the limited data available. We evaluate the achievable accuracy through experiments on an inexpensive 6-DoF robot.

Robotic manipulation requires explicit or implicit knowledge of the robot's joint positions.