Robotics & Embodied AI - Weekly Roundup

It is found that the best frozen text encoder is simply the one whose geometry best matches the video features, and the best frozen text encoder is simply the one whose geometry best matches the video features.

MILER, an end-to-end policy framework with zero-shot sim-to-real transfer of both perception and control, is presented and extensively evaluated on a diverse test track comprising numerous challenges.

Imagine-TAMP is introduced, an interleaved planning and execution framework that uses semantic and geometric imagination to compare alternative task-level strategies under partial observability before committing to expensive motion planning and experiments show that imagination-guided evaluation improves observation-versus-manipulation decisions.

Comparisons support complementary benefits of feature-wise gated tactile fusion and training-only predictive supervision of TacSushi, a tactile-grounded, Cosmos3-based world-action policy that learns from recorded future consequences while acting on current observations.

MILER, an end-to-end policy framework with zero-shot sim-to-real transfer of both perception and control, is presented and extensively evaluated on a diverse test track comprising numerous challenges.

Imagine-TAMP is introduced, an interleaved planning and execution framework that uses semantic and geometric imagination to compare alternative task-level strategies under partial observability before committing to expensive motion planning and experiments show that imagination-guided evaluation improves observation-versus-manipulation decisions.

Comparisons support complementary benefits of feature-wise gated tactile fusion and training-only predictive supervision of TacSushi, a tactile-grounded, Cosmos3-based world-action policy that learns from recorded future consequences while acting on current observations.

This work introduces LIFD (Look, Imagine, Focus, and Do), a framework for persistent, 3D-aware scene memory that learns a scene-token representation from multi-view agreement and completes it from a single RGB view and recurrent memory.

This work proposes Set of Centre Active Receptive Fields (SCARF), a real-time velocity-invariant representation that preserves raw events while consistently handling fast motion, stationary scenes, and independently moving objects and achieves state-of-the-art performance in both computational efficiency and representation quality.

Sampling-Guided Policy Search (SGPS), which couples recurring action-target refinement by sampling-based model-predictive control with first-order policy optimization with first-order policy optimization, and shows that refinement improves policy learning beyond initialization and tracking alone.

This paper introduces multi-horizon multimodal prediction in ABBR, an agile tactile World Action Model for contact-rich robot control, which provides supervision for visual latent at a larger temporal offset while predicting the tactile latent in the next frame to capture fine-grained contact dynamics.

Delphi Scanner is introduced, a static malware detection system for Windows PE files that balances efficiency with behavioral interpretation that uses a convolutional neural network to model Windows API sequences to classify PE and a decoupled interpretation layer based on a rule-based layer to categorize APIs into high-level malicious capabilities.

HIL-UMI is introduced, a policy-guided Universal Manipulation Interface (UMI) framework for robot-free human-in-the-loop VLA post-training that preserves the iterative and policy-aware nature of human-in-the-loop learning while decoupling data collection from robot deployment.

Experiments on a real robot platform demonstrate that MaskHarness-WAM substantially outperforms limited-horizon policies on sequential multi-object manipulation, showing its effectiveness in extending local manipulation skills to reliable long-horizon execution.

This paper presents JANUS, a novel targeted denial-of-service attack against beam-hopping systems in LEO networks, and shows that a small botnet of compromised terminals can inject legitimate user traffic into carefully selected non-victim cells to manipulate the beam-hopping scheduler's view of demand.

A lightweight single-probe audit is designed that applies a controlled lengthening intervention under PTIA, and achieves an average detection rate of 85.1% with false-positive rates below 2%.

This work introduces DexTouch-WM, an action-conditioned world model that learns from scalable human touch to jointly predict future RGB observations and bilateral tactile dynamics, and shows that scalable human interaction provides a complementary data axis for learning dexterous robot world models.

OmniMimic, a training framework that turns directionally limited animal demonstrations into a single multi-gait policy over target per-axis velocity ranges, and uses a shared actor with soft-gated, gait-specialized residual experts to balance reusable locomotion skills with gait-specific corrections.

It is revealed that 2D image and 3D shape encoders exhibit largely disjoint failure patterns and rarely share identical wrong labels, whereas two 2D image encoders frequently repeat the same errors, which makes the 2D and 3D pair inherently complementary.

A unified four-step post-training pipeline comprising a learned temporal hand-action codec, supervised fine-tuning, DAgger, and real-world residual reinforcement learning provides a practical path for adapting VLA foundation models to reliable real-world dexterous manipulation.

Results show that, for proximity-defined manipulation hazards, margin selection and fallback strategy can matter more than predicate structure in determining the safety--utility trade-off of a runtime filter.

This work presents Co-VLA, which applies consensus optimization using the Alternating Direction Method of Multipliers~(ADMM) to federated VLA training, and demonstrates that Co-VLA achieves performance comparable to centralized training in both full-model training and parameter-efficient fine-tuning settings.

RotateIt!, a single-arm framework that uses adaptive axial rotation for dynamic garment unfolding, is presented, to the best of the knowledge, it is the first unfolding framework to employ dynamic axial rotation as its primary manipulation primitive.

SLAMSqueezeBench, a framework that allows testing of SLAM systems under realistic workloads on edge hardware, and compares nine SLAM systems spanning classical systems, learning-based systems, and approaches for Gaussian splatting.

The concept of traversability is extended from robot-centric to human-centric by accounting for the operator's cognitive states, such as attention, workload, and risk tolerance or awareness, termed tele-traversability, and is revisited and extended to teleoperation settings.

This work presents RebarSim, a visual sim-to-real system trained entirely in simulation, a privileged state-based teacher is trained with reinforcement learning over procedurally generated rebar geometries, then distilled into a multi-view student that maps raw RGB and proprioception directly to actions under extensive domain randomization.

SkipVLA is a hybrid policy that combines a pretrained VLA with a classical motion planner, using the planner for free-space motion and querying the VLA only for contact-rich skills such as grasping and placing, demonstrating up to 2.5x faster task completion and significantly lower energy consumption.

GeoAAC is proposed, a geometry-based adaptive action chunking method for flow-based VLA policies that adjusts the action horizon according to the reliability of the current action prediction, with geometric variation across action prefixes remaining positively correlated with predictive uncertainty.

SafeHarness is presented, which equips the model with two obstacle-aware harnesses that enable it to prioritize the safety constraint, and surpasses the previous SOTA by 6.5% and 27.0%, respectively.

MAGMA-GEN is an on-policy data-generation pipeline that converts ambiguous failed rollouts into validated recovery supervision and improves task success and recovery capabilities, against distillation and trajectory-repair baselines under evolving task constraints in both simulation and real-robot execution.

This work presents TouchSight, a monocular egocentric vision framework for dense full-hand contact force prediction that leverages 500 hours of pressure-glove recordings and extensive hand-object interaction (HOI) data and demonstrates that dense tactile signals can be recovered from egocentric vision alone, without tactile instrumentation at capture time.

V2-STRep is presented, a zero-shot framework that converts generated video motion into reusable robot skills through VLM-grounded structured task representations, and demonstrates improved execution success over baselines, reliable cross-scene transfer of successfully acquired skills, and adaptation to changed deployment instructions.

Private bot-checking behavior from the demand side constitutes a mass, distributed form of platform auditing that tracks meaningful signals of automation, establishing audit-tool query logs as a novel lens on public responses to platform manipulation.

An attack space for MEV on DAG-based BFT protocols is presented, organized around four families: the adversary, the protocol, the target, and the deployment, and it is revealed that every protocol evaluated is vulnerable to at least a subset of the MEV attacks in this space.

AnyviewMeter is introduced, a geometry-conditioned adaptation framework for robotic reward models that represent task progress as a scalar reward signal that combines low-rank fine-tuning with token-aligned Plucker rays and synchronous block attention.

TADreamer is presented, a zero-shot framework that grounds video-imagined navigation in measured geometry without task-specific training or fine-tuning, and reduces mean absolute depth error and mean absolute relative depth error compared with NavDreamer.

Experiments demonstrate that MoWAM achieves strong in-distribution performance, improved out-of-distribution robustness, and higher average real-world success than representative WAM baselines, demonstrating that explicit future motion provides an effective and efficient basis for inference-time scaling.

A task-semantic scene-grounded framework for learning roofer-style whole-body motions on a Unitree G1, establishing scene-grounded human motion learning as a promising basis for construction-oriented humanoid motion primitives.

The collected dataset represents the first multimodal dataset that provides continuous visual, physiological, and operator's bilateral motion tracking of temporal affective state shifts during bimanual teleoperation and outperforms zero-shot baselines (Qwen, MiniCPM-V in tracking temporal state dynamics).

This work shows that view-invariant control can be achieved without explicit test-time 3D sensing by encoding geometric knowledge into the visual representation during simulation, and presents AnyViewDex, an asymmetric training pipeline that combines multi-view contrastive alignment with privileged 3D geometric supervision.

This work examines the trade-off between explicitly incorporating compliance through series-elastic actuation and using so-called proprioceptive, low-gear-ratio transmissions, and explores how these mechanisms interact with control policies during locomotion.

WorldContact is presented, a contact-centric world model for deformable-object manipulation, constructed from a limited set of high-quality trajectories to generate additional training data efficiently and fine-tune an existing vision-language-action policy and deploy it directly on a real robot.

The results show that a diffusion policy can benefit substantially from knowing where an interaction is heading, without being told exactly where to move.

This work investigates compositional reuse at two levels: within a subtask, combining familiar entities, action types, and action modifiers; and across subtasks, reusing learned subtasks in unseen long-horizon tasks, which validate the effectiveness of the method for instruction-dependent generalization at both levels.

ContrAgent is presented, a contract-based framework for symbolic temporal supervision of LLM agents that matches state-of-the-art LLM-judge and rule-based guardrail baselines while producing deterministic, reproducible verdicts and, in the online mode, orders-of-magnitude lower per-call latency.

WAVE-Go, an image-goal navigation framework that separates world-action prediction from interruptible command execution, and its executor adaptively selects an action prefix and cancels pending commands when updated observations invalidate execution, support adaptive, interruptible execution as a means of balancing navigation performance and planning overhead.

The Decoupled Embodiment Model (DEM), which pairs a fine-tuned DINOv3 encoder and a frozen NeoBERT encoder with a MeanFlow head that generates each action chunk in a single forward pass, achieves observed success comparable to state-of-the-art VLM-backbone policies under the evaluation protocol.

This work instantiates Real-Time EXPO-FT, an RL framework for finetuning real-time VLA policies that meets the real-time control requirements of dynamic real-world manipulation, demonstrating rapid, sample-efficient adaptation to challenging real-world dynamics.

The results show that replay retention depends on change magnitude and on how the dynamics evolve, and an estimator built from interaction data can still provide the quantities needed to choose a replay strategy after permanent changes.

A metric interaction framework that models object-level and scene-level interactions in physical Cartesian space at a shared metric scale and improves diverse VLA and WAM baselines with a small number of additional parameters and training steps is introduced.

This work introduces market signal injection (MSI), an attack that manipulates numerical formatting, competitor ordering, or qualitative market commentary without issuing explicit instructions to motivate defenses that account for interactions among agents.

Results show ARC can create and support rural robotics programs, train undergraduate AI and robotics mentors, and potentially scale mentorship across a region, as well as create a spatial Markov model of ARC's growth and simulate it using the state of Indiana as a testbed.

The results show that work-aware policy optimization can suppress physically inefficient motion without requiring an explicit differentiable dynamics model.

A function-preserving Real-to-Sim-to-Real framework that generates synthetic demonstrations from reconstructed assets without teleoperated source trajectories is proposed, enabling robust generalization across unseen object geometries and diverse visual conditions in contact-rich and long-horizon tasks.

ForceDelta-VLA is presented, a correction-distillation framework that constructs an explicit force-correction target using paired predictions from a frozen teacher's force-conditioned and learned force-agnostic modes, and reduces mean peak contact force over successful trials by approximately 26% on both platforms.

RAFAIL is introduced, a framework for detecting execution failures during robotic manipulation that requires no failure data and achieves 73.4% balanced accuracy across three real-world robotic manipulation tasks, outperforming the strongest evaluated OOD- and uncertainty-based baselines.

A synthetic-to-real framework coupling physically validated data generation with a hierarchical, diffusion-based controller is proposed, which outperforms state-of-the-art manipulation policies in transport smoothness and dynamic stability.

UMI-Bridge, which uses UMI as an intermediate domain to align representations according to action equivalence rather than pixel similarity, supports action-anchored latent alignment for data-efficient robot learning and UMI-to-robot task transfer.

CANTABILE is proposed, a dynamics-aware framework for robotic piano performance that closes the score-to-contact loop by conditioning the policy on upcoming velocity goals and mapping each key's angular velocity at onset back to MIDI velocity, and enables runtime control of performance intensity without retraining.

This work studies open-vocabulary task-oriented dexterous grasp generation, where a robot must infer functional intent from free-form language, ground it in multi-view visual observations and object geometry, and generate an executable high-degree-of-freedom grasp.

Experiments show that WholeBodyWAM consistently benefits from increased motion-pretraining scale, improves future-motion prediction and downstream task performance, and transfers effectively to real-world humanoid manipulation.

The proposed method accelerates training and enables stable policies where matched raw-action PPO remains near failure, with successful sim-to-real transfer in in-hand and arm-hand tasks.

A training-free framework that reformulates failure recovery as test-time scaling over causal histories, which decomposes recovery into three coupled decisions: when to revise the causal history, where to recover a reliable history prefix, and which history configuration best supports subsequent execution.

This work investigates an alternative representation in which each joint trajectory is expressed as a command-conditioned Fourier series and modified online using feedback from the robot state, showing periodic structure without an explicit Fourier generator.

GPT-Policy is introduced, a general-agent framework for in-context robot learning that integrates a context compiler that preserves task-relevant visual transitions, a VLM that proposes robot-tool actions, and a constrained controller that verifies and executes each action and reports its outcome.

This work proposes Collision Mesh Poisoning (CMP), the first poisoning attack against robotic manipulation delivered through the 3D asset supply chain, and evaluates several defenses and shows that they are insufficient to defend against CMP, highlighting the need for new defenses.

It is found that the best frozen text encoder is simply the one whose geometry best matches the video features, and the best frozen text encoder is simply the one whose geometry best matches the video features.

ParticleSplat is presented, a self-supervised object-centric learning method that decomposes scenes into a set of latent particles representing semantic entities through feedforward 3D Gaussian Splatting, and it is established that the learned 3D representation improves downstream performance on robotic manipulation tasks.

WholeBodyWAM is introduced, which jointly predicts future visual dynamics, manipulation actions, and whole-body control intents for generalizable humanoid loco-manipulation and preserves pre-trained world-action priors while grounding heterogeneous whole-body controller (WBC) semantics and coordinating whole-body behavior.

PhysStream is an autoregressive model for physics-grounded image-to-video synthesis that incorporates structured scene memory and supports fine-grained motion control via sparse velocity-increment signals that encode physical quantities, letting the model learn the underlying dynamics.

Latent Chip Adaptation from Probes (LCAP), a population-informed framework that decomposes hardware adaptation into a transferable population correction and probe-inferred latent personalization, is proposed.

The original ACT paper reported that encoder removal dropped the mean success rate from 35% to 2% on two simulated tasks with human demonstrations, but this drop does not reappear in tests, although smaller gains or losses in success rate remain uncertain.

Weave is presented, a unified framework for learning whole-body dexterous humanoid-object interaction from captured human demonstrations that converts captured human-object interactions into executable robot-object references through contact-aware retargeting and approach-motion completion.

This framework surveys manipulation, navigation, locomotion, autonomous driving, and general embodied learning, tracing technical progressions, clarifying capability requirements, and examining datasets, benchmarks, and evaluation protocols.

This article presents SEMA-GUARD, a framework that uses semantic analysis and graph neural networks to identify flaws in assembly code, and results imply that including semantic information in graph-based models may be a successful method for identifying vulnerabilities in compiled code.

Ex vivo central-airway obstruction experiments and simulations on deforming cube models demonstrate reliable and accurate reconstruction under both rigid motion and deformation, highlighting the potential of this probabilistic approach for principled, measurement-driven mesh state estimation in deformable object reconstruction.