VLA 研究日報VLA 研究日报

Vision-Language-Action (VLA) models have advanced robotic manipulation by combining vision, language, and proprioception to predict actions. However, previous methods fuse proprioceptive signals directly with vision-language features, resulting in state-dominant bias and \textbf{false completions} despite visible execution failures. We systematically analyze this failure mode, attributing it to modality imbalance, where policies overly rely on internal state progression and underuse visual evidence. To address this, we introduce the first \textbf{False-Completion Benchmark Suite}, featuring eight tasks with three controlled perturbations (\emph{Object Drop}, \emph{Distractor Swap}, \emph{Relayout}) to comprehensively evaluate false completion. Moreover, we propose \textbf{ReViP}, a novel VLA framework with \textbf{Vi}sion-\textbf{P}roprioception \textbf{Re}balance to enhance visual grounding and robustness under perturbations. The key insight is to introduce auxiliary \emph{progress-aware visual cues} to adaptively modulate the coupling between semantic perception and proprioceptive dynamics. Specifically, progress-aware visual cues are extracted by an external Task-Stage Observer, which performs task-relevant reasoning on real-time observations to drive task-stage feature-wise linear modulation, enhancing environmental awareness and mitigating state-driven errors. Extensive experiments show that ReViP effectively mitigates false completion and improves success rates over strong VLA baselines, achieving a \textbf{26\%} gain over $π_0$ model on our suite, with gains extending to LIBERO, RoboTwin 2.0, and real-world evaluations.

Vision-Language-Action (VLA) models are dominant in embodied intelligence but are constrained by inference overheads. While model quantization alleviates these bottlenecks for edge deployment, static quantization approaches remain suboptimal for VLAs due to two critical challenges: (1) Temporal-dynamic sensitivity, where fixed precision wastes resources by ignoring stage-varying error tolerances; and (2) Real-time allocation, where identifying real-time sensitivity to guide bit allocation remains unsolved. To address these challenges, we propose DyQ-VLA, a dynamic quantization framework for VLAs. Specifically, a sensitivity-aware switching strategy leverages real-time kinematic proxies to trigger the bit-width switch, while a kinematic-guided module dynamically allocates the optimal bit-width. Experiments show that DyQ-VLA requires only 30.9% of the original memory footprint while maintaining 99.5% of its original performance, achieving 1.49x simulation and up to 1.43x real-world speedups.

Recent advances in Visual-Language-Action (VLA) models have shown promising potential for robotic manipulation tasks. However, real-world robotic tasks often involve long-horizon, multi-step problem-solving and require generalization for continual skill acquisition, extending beyond single actions or skills. These challenges present significant barriers for existing VLA models, which use monolithic action decoders trained on aggregated data, resulting in poor scalability. To address these challenges, we propose AtomicVLA, a unified planning-and-execution framework that jointly generates task-level plans, atomic skill abstractions, and fine-grained actions. AtomicVLA constructs a scalable atomic skill library through a Skill-Guided Mixture-of-Experts (SG-MoE), where each expert specializes in mastering generic yet precise atomic skills. Furthermore, we introduce a flexible routing encoder that automatically assigns dedicated atomic experts to new skills, enabling continual learning. We validate our approach through extensive experiments. In simulation, AtomicVLA outperforms $π_{0}$ by 2.4\% on LIBERO, 10\% on LIBERO-LONG, and outperforms $π_{0}$ and $π_{0.5}$ by 0.22 and 0.25 in average task length on CALVIN. Additionally, our AtomicVLA consistently surpasses baselines by 18.3\% and 21\% in real-world long-horizon tasks and continual learning. These results highlight the effectiveness of atomic skill abstraction and dynamic expert composition for long-horizon and lifelong robotic tasks. The project page is \href{https://zhanglk9.github.io/atomicvla-web/}{here}.

Recent vision-language-action (VLA) models have significantly advanced robotic manipulation by unifying perception, reasoning, and control. To achieve such integration, recent studies adopt a predictive paradigm that models future visual states or world knowledge to guide action generation. However, these models emphasize forecasting outcomes rather than reasoning about the underlying process of change, which is essential for determining how to act. To address this, we propose $Δ$VLA, a prior-guided framework that models world-knowledge variations relative to an explicit current-world knowledge prior for action generation, rather than regressing absolute future world states. Specifically, 1) to construct the current world knowledge prior, we propose the Prior-Guided WorldKnowledge Extractor (PWKE). It extracts manipulable regions, spatial relations, and semantic cues from the visual input, guided by auxiliary heads and prior pseudo labels, thus reducing redundancy. 2) Building upon this, to represent how world knowledge evolves under actions, we introduce the Latent World Variation Quantization (LWVQ). It learns a discrete latent space via a VQ-VAE objective to encode world knowledge variations, shifting prediction from full modalities to compact latent. 3)Moreover, to mitigate interference during variation modeling, we design the Conditional Variation Attention (CV-Atten), whichpromotes disentangled learning and preserves the independence of knowledge representations. Extensive experiments on both simulated benchmarks and real-world robotic tasks demonstrate $Δ$VLA achieves state-of-the-art performance while improving efficiency. Code and real-world execution videos are available at https://github.com/JiuTian-VL/DeltaVLA.

Diffusion and flow matching policies offer expressive, multimodal action modeling, yet they are frequently unstable in online reinforcement learning (RL) due to intractable likelihoods and gradients propagating through long sampling chains. Conversely, tractable parameterizations such as Gaussians lack the expressiveness needed for complex control -- exposing a persistent tension between optimization stability and representational power. We address this tension with a key structural principle: decoupling optimization from generation. Building on this, we introduce GoRL (Generative Online Reinforcement Learning), an algorithm-agnostic framework that trains expressive policies from scratch by confining policy optimization to a tractable latent space while delegating action synthesis to a conditional generative decoder. Using a two-timescale alternating schedule and anchoring decoder refinement to a fixed prior, GoRL enables stable optimization while continuously expanding expressiveness. Empirically, GoRL consistently outperforms unimodal and generative baselines across diverse continuous-control tasks. Notably, on the challenging HopperStand task, it achieves episodic returns exceeding 870 -- more than $3\times$ that of the strongest baseline -- demonstrating a practical path to policies that are both stable and highly expressive. Our code is publicly available at https://github.com/bennidict23/GoRL.

Robotic systems operating in unstructured environments must operate under significant uncertainty arising from intermittent contacts, frictional variability, and unmodeled compliance. While recent model-free approaches have demonstrated impressive performance, many deployment settings still require predictive models that support planning, constraint handling, and online adaptation. Analytical rigid-body models provide strong physical structure but often fail to capture complex interaction effects, whereas purely data-driven models may violate physical consistency, exhibit data bias, and accumulate long-horizon drift. In this work, we propose STRIDE, a dynamics learning framework that explicitly separates conservative rigid-body mechanics from uncertain, effectively stochastic non-conservative interaction effects. The structured component is modeled using a Lagrangian Neural Network (LNN) to preserve energy-consistent inertial dynamics, while residual interaction forces are represented using Conditional Flow Matching (CFM) to capture multi-modal interaction phenomena. The two components are trained jointly end-to-end, enabling the model to retain physical structure while representing complex stochastic behavior. We evaluate STRIDE on systems of increasing complexity, including a pendulum, the Unitree Go1 quadruped, and the Unitree G1 humanoid. Results show 20% reduction in long-horizon prediction error and 30% reduction in contact force prediction error compared to deterministic residual baselines, supporting more reliable model-based control in uncertain robotic environments.

When an RL agent's observations are gradually corrupted, at what drift rate does it "wake up" -- and what determines this boundary? We study world model-based self-monitoring under continuous observation drift across four MuJoCo environments, three detector families (z-score, variance, percentile), and three model capacities. We find that (1) a sharp detection threshold $\varepsilon^*$ exists universally: below it, drift is absorbed as normal variation; above it, detection occurs rapidly. The threshold's existence and sigmoid shape are invariant across all detector families and model capacities, though its position depends on the interaction between detector sensitivity, noise floor structure, and environment dynamics. (2) Sinusoidal drift is completely undetectable by all detector families -- including variance and percentile detectors with no temporal smoothing -- establishing this as a world model property rather than a detector artifact. (3) Within each environment, $\varepsilon^*$ follows a power law in detector parameters ($R^2 = 0.89$-$0.97$), but cross-environment prediction fails ($R^2 = 0.45$), revealing that the missing variable is environment-specific dynamics structure $\partial \mathrm{PE}/\partial\varepsilon$. (4) In fragile environments, agents collapse before any detector can fire ("collapse before awareness"), creating a fundamentally unmonitorable failure mode. Our results reframe $\varepsilon^*$ from an emergent world model property to a three-way interaction between noise floor, detector, and environment dynamics, providing a more defensible and empirically grounded account of self-monitoring boundaries in RL agents.

Tactile sensation is essential for contact-rich manipulation tasks. It provides direct feedback on object geometry, surface properties, and interaction forces, enhancing perception and enabling fine-grained control. An inherent limitation of tactile sensors is that readings are available only when an object is touched. This precludes their use during planning and the initial execution phase of a task. Predicting tactile information from visual information can bridge this gap. A common approach is to learn a direct mapping from camera images to the output of vision-based tactile sensors. However, the resulting model will depend strongly on the specific setup and on how well the camera can capture the area where an object is touched. In this work, we introduce FlowTouch, a novel model for view-invariant visuo-tactile prediction. Our key idea is to use an object's local 3D mesh to encode rich information for predicting tactile patterns while abstracting away from scene-dependent details. FlowTouch integrates scene reconstruction and Flow Matching-based models for image generation. Our results show that FlowTouch is able to bridge the sim-to-real gap and generalize to new sensor instances. We further show that the resulting tactile images can be used for downstream grasp stability prediction. Our code, datasets and videos are available at https://flowtouch.github.io/

Model-based reinforcement learning (MBRL) agents operating in high-dimensional observation spaces, such as Dreamer, rely on learning abstract representations for effective planning and control. Existing approaches typically employ reconstruction-based objectives in the observation space, which can render representations sensitive to task-irrelevant details. Recent alternatives trade reconstruction for auxiliary action prediction heads or view augmentation strategies, but perform worse in the Crafter environment than reconstruction-based methods. We close this gap between Dreamer and reconstruction-free models by introducing a JEPA-style predictor defined on continuous, deterministic representations. Our method matches Dreamer's performance on Crafter, demonstrating effective world model learning on this benchmark without reconstruction objectives.

Our aim is to learn to solve long-horizon decision-making problems in complex robotics domains given low-level skills and a handful of short-horizon demonstrations containing sequences of images. To this end, we focus on learning abstract symbolic world models that facilitate zero-shot generalization to novel goals via planning. A critical component of such models is the set of symbolic predicates that define properties of and relationships between objects. In this work, we leverage pretrained vision-language models (VLMs) to propose a large set of visual predicates potentially relevant for decision-making, and to evaluate those predicates directly from camera images. At training time, we pass the proposed predicates and demonstrations into an optimization-based model-learning algorithm to obtain an abstract symbolic world model that is defined in terms of a compact subset of the proposed predicates. At test time, given a novel goal in a novel setting, we use the VLM to construct a symbolic description of the current world state, and then use a search-based planning algorithm to find a sequence of low-level skills that achieves the goal. We demonstrate empirically across experiments in both simulation and the real world that our method can generalize aggressively, applying its learned world model to solve problems with a wide variety of object types, arrangements, numbers of objects, and visual backgrounds, as well as novel goals and much longer horizons than those seen at training time.

[MIT]

Tactile sensing provides local essential information that is complementary to visual perception, such as texture, compliance, and force. Despite recent advances in visuotactile representation learning, challenges remain in fusing these modalities and generalizing across tasks and environments without heavy reliance on pre-trained vision-language models. Moreover, existing methods do not study positional encodings, thereby overlooking the multi-stage spatial reasoning needed to capture fine-grained visuotactile correlations. We introduce ViTaPEs, a transformer-based architecture for learning task-agnostic visuotactile representations from paired vision and tactile inputs. Our key idea is a two-stage positional injection: local (modality-specific) positional encodings are added within each stream, and a global positional encoding is added on the joint token sequence immediately before attention, providing a shared positional vocabulary at the stage where cross-modal interaction occurs. We make the positional injection points explicit and conduct controlled ablations that isolate their effect before a token-wise nonlinearity versus immediately before self-attention. Experiments on multiple large-scale real-world datasets show that ViTaPEs not only surpasses state-of-the-art baselines across various recognition tasks but also demonstrates zero-shot generalization to unseen, out-of-domain scenarios. We further demonstrate the transfer-learning strength of ViTaPEs in a robotic grasping task, where it outperforms state-of-the-art baselines in predicting grasp success. Project page: https://sites.google.com/view/vitapes

Imitation learning for manipulation has a well-known data scarcity problem. Unlike natural language and 2D computer vision, there is no Internet-scale corpus of data for dexterous manipulation. One appealing option is egocentric human video, a passively scalable data source. However, existing large-scale datasets such as Ego4D do not have native hand pose annotations and do not focus on object manipulation. To this end, we use Apple Vision Pro to collect EgoDex: the largest and most diverse dataset of dexterous human manipulation to date. EgoDex has 829 hours of egocentric video with paired 3D hand and finger tracking data collected at the time of recording, where multiple calibrated cameras and on-device SLAM can be used to precisely track the pose of every joint of each hand. The dataset covers a wide range of diverse manipulation behaviors with everyday household objects in 194 different tabletop tasks ranging from tying shoelaces to folding laundry. Furthermore, we train and systematically evaluate imitation learning policies for hand trajectory prediction on the dataset, introducing metrics and benchmarks for measuring progress in this increasingly important area. By releasing this large-scale dataset, we hope to push the frontier of robotics, computer vision, and foundation models. EgoDex is publicly available for download at https://github.com/apple/ml-egodex.

[Berkeley]

Action-conditioned video prediction models (often referred to as world models) have shown strong potential for robotics applications, but existing approaches are often slow and struggle to capture physically consistent interactions over long horizons, limiting their usefulness for scalable robot policy training and evaluation. We present Interactive World Simulator, a framework for building interactive world models from a moderate-sized robot interaction dataset. Our approach leverages consistency models for both image decoding and latent-space dynamics prediction, enabling fast and stable simulation of physical interactions. In our experiments, the learned world models produce interaction-consistent pixel-level predictions and support stable long-horizon interactions for more than 10 minutes at 15 FPS on a single RTX 4090 GPU. Our framework enables scalable demonstration collection solely within the world models to train state-of-the-art imitation policies. Through extensive real-world evaluation across diverse tasks involving rigid objects, deformable objects, object piles, and their interactions, we find that policies trained on world-model-generated data perform comparably to those trained on the same amount of real-world data. Additionally, we evaluate policies both within the world models and in the real world across diverse tasks, and observe a strong correlation between simulated and real-world performance. Together, these results establish the Interactive World Simulator as a stable and physically consistent surrogate for scalable robotic data generation and faithful, reproducible policy evaluation.

We introduce SPIRAL, a self-improving planning and iterative reflective action world modeling closed-loop framework that enables controllable long-horizon video generation conditioned on high-level semantic actions. Existing one-shot video generation models operate in open-loop, often resulting in incomplete action execution, weak semantic grounding, and temporal drift. SPIRAL formulates ActWM as a closed-loop think-act-reflect process, where generation proceeds step by step under explicit planning and feedback. A PlanAgent decomposes abstract actions into object-centric sub-actions, while a CriticAgent evaluates intermediate results and guides iterative refinement with long-horizon memory. This closed-loop design naturally supports RL evolving optimization, improving semantic alignment and temporal consistency over extended horizons. We further introduce the ActWM-Dataset and ActWM-Bench for training and evaluation. Experiments across multiple TI2V backbones demonstrate consistent gains on ActWM-Bench and mainstream video generation benchmarks, validating SPIRAL's effectiveness.

World models enable planning in imagined future predicted space, offering a promising framework for embodied navigation. However, existing navigation world models often lack action-conditioned consistency, so visually plausible predictions can still drift under multi-step rollout and degrade planning. Moreover, efficient deployment requires few-step diffusion inference, but existing distillation methods do not explicitly preserve rollout consistency, creating a training-inference mismatch. To address these challenges, we propose MWM, a mobile world model for planning-based image-goal navigation. Specifically, we introduce a two-stage training framework that combines structure pretraining with Action-Conditioned Consistency (ACC) post-training to improve action-conditioned rollout consistency. We further introduce Inference-Consistent State Distillation (ICSD) for few-step diffusion distillation with improved rollout consistency. Our experiments on benchmark and real-world tasks demonstrate consistent gains in visual fidelity, trajectory accuracy, planning success, and inference efficiency. Code: https://github.com/AIGeeksGroup/MWM. Website: https://aigeeksgroup.github.io/MWM.

Learned world models excel at interpolative generalization but fail at extrapolative generalization to novel physical properties. This limitation arises because they learn statistical correlations rather than the environment's underlying generative rules, such as physical invariances and conservation laws. We argue that learning these invariances is key to robust extrapolation. To achieve this, we first introduce \textbf{Symmetry Exploration}, an unsupervised exploration strategy where an agent is intrinsically motivated by a Hamiltonian-based curiosity bonus to actively probe and challenge its understanding of conservation laws, thereby collecting physically informative data. Second, we design a Hamiltonian-based world model that learns from the collected data, using a novel self-supervised contrastive objective to identify the invariant physical state from raw, view-dependent pixel observations. Our framework, \textbf{DreamSAC}, trained on this actively curated data, significantly outperforms state-of-the-art baselines in 3D physics simulations on tasks requiring extrapolation.

Learning robotic manipulation policies through supervised learning from demonstrations remains challenging when policies encounter execution variations not explicitly covered during training. While incorporating historical context through attention mechanisms can improve robustness, standard approaches process all past states in a sequence without explicitly modeling the temporal structure that demonstrations may include, such as failure and recovery patterns. We propose a Cross-State Transition Attention Transformer that employs a novel State Transition Attention (STA) mechanism to modulate standard attention weights based on learned state evolution patterns, enabling policies to better adapt their behavior based on execution history. Our approach combines this structured attention with temporal masking during training, where visual information is randomly removed from recent timesteps to encourage temporal reasoning from historical context. Evaluation in simulation shows that STA consistently outperforms standard attention approach and temporal modeling methods like TCN and LSTM networks, achieving more than 2x improvement over cross-attention on precision-critical tasks. The source code and data can be accessed at https://github.com/iit-DLSLab/croSTAta

[IIT]

We propose DemoDiffusion, a simple method for enabling robots to perform manipulation tasks by imitating a single human demonstration, without requiring task-specific training or paired human-robot data. Our approach is based on two insights. First, the hand motion in a human demonstration provides a useful prior for the robot's end-effector trajectory, which we can convert into a rough open-loop robot motion trajectory via kinematic retargeting. Second, while this retargeted motion captures the overall structure of the task, it may not align well with plausible robot actions in-context. To address this, we leverage a pre-trained generalist diffusion policy to modify the trajectory, ensuring it both follows the human motion and remains within the distribution of plausible robot actions. Unlike approaches based on online reinforcement learning or paired human-robot data, our method enables robust adaptation to new tasks and scenes with minimal effort. In real-world experiments across 8 diverse manipulation tasks, DemoDiffusion achieves 83.8\% average success rate, compared to 13.8\% for the pre-trained policy and 52.5\% for kinematic retargeting, succeeding even on tasks where the pre-trained generalist policy fails entirely. Project page: https://demodiffusion.github.io/

[CMU]

Learning expressive and efficient policy functions is a promising direction in reinforcement learning (RL). While flow-based policies have recently proven effective in modeling complex action distributions with a fast deterministic sampling process, they still face a trade-off between expressiveness and computational burden, which is typically controlled by the number of flow steps. In this work, we propose mean velocity policy (MVP), a new generative policy function that models the mean velocity field to achieve the fastest one-step action generation. To ensure its high expressiveness, an instantaneous velocity constraint (IVC) is introduced on the mean velocity field during training. We theoretically prove that this design explicitly serves as a crucial boundary condition, thereby improving learning accuracy and enhancing policy expressiveness. Empirically, our MVP achieves state-of-the-art success rates across several challenging robotic manipulation tasks from Robomimic and OGBench. It also delivers substantial improvements in training and inference speed over existing flow-based policy baselines.

We revisit Vision-Language-Action through a neuroscience-inspired triad. Biologically, the Cerebrum provides stable high-level multimodal priors and remains frozen; the Pons Adapter integrates these cortical features with real-time proprioceptive inputs and compiles intent into execution-ready tokens; and the Cerebellum (ParaCAT) performs fast, parallel categorical decoding for online control, with hysteresis/EMA/temperature/entropy for stability. A fixed-ratio schedule and two-stage feature caching make the system compute-aware and reproducible. Inspired by active, foveated vision, our wrist ROIs are geometrically tied to the end-effector via calibrated projection, providing a movement-stabilized, high-resolution view that is sensitive to fine-grained pose changes and complements the global context of the main view. The design is modular: upgrading the Cerebrum only retrains the Pons; changing robots only trains the Cerebellum; cerebellum-only RL can further refine control without touching high-level semantics. As a concept-and-protocol paper with preliminary evidence, we outline a timing protocol under matched conditions (GPU, resolution, batch) to verify anticipated efficiency gains. We also report preliminary LIBERO evidence showing that split feature caching reduces training time (7.5h to 4.5h) and improves average success (86.5% to 92.5%) under official N1.5 head-only training, and that SaiVLA0 reaches 99.0% mean success.

[ASU|Huang]

Mobile manipulation requires the coordinated control of a mobile base and a robotic arm while simultaneously perceiving both global scene context and fine-grained object details. Existing single-view approaches often fail in unstructured environments due to limited fields of view, exploration, and generalization abilities. Moreover, classical controllers, although stable, struggle with efficiency and manipulability near singularities. To address these challenges, we propose M4Diffuser, a hybrid framework that integrates a Multi-View Diffusion Policy with a novel Reduced and Manipulability-aware QP (ReM-QP) controller for mobile manipulation. The diffusion policy leverages proprioceptive states and complementary camera perspectives with both close-range object details and global scene context to generate task-relevant end-effector goals in the world frame. These high-level goals are then executed by the ReM-QP controller, which eliminates slack variables for computational efficiency and incorporates manipulability-aware preferences for robustness near singularities. Comprehensive experiments in simulation and real-world environments show that M4Diffuser achieves 7 to 56 percent higher success rates and reduces collisions by 3 to 31 percent over baselines. Our approach demonstrates robust performance for smooth whole-body coordination, and strong generalization to unseen tasks, paving the way for reliable mobile manipulation in unstructured environments. Details of the demo and supplemental material are available on our project website https://sites.google.com/view/m4diffuser.

We present generative predictive control (GPC), a learning control framework that (i) clones a generative diffusion-based policy from expert demonstrations, (ii) trains a predictive action-conditioned world model from both expert demonstrations and random explorations, and (iii) synthesizes an online planner that ranks and optimizes the action proposals from (i) by looking ahead into the future using the world model from (ii). Across a variety of robotic manipulation tasks, we demonstrate that GPC consistently outperforms behavior cloning in both state-based and vision-based settings, in simulation and in the real world.

[Harvard]

World models, which predict future transitions from past observation and action sequences, have shown great promise for improving data efficiency in sequential decision-making. However, existing world models often require extensive domain-specific training and still produce low-fidelity, coarse predictions, limiting their usefulness in complex environments. In contrast, video diffusion models trained on large-scale internet data have demonstrated impressive capabilities in generating high-quality videos that capture diverse real-world dynamics. In this work, we present Vid2World, a general approach for leveraging and transferring pre-trained video diffusion models into interactive world models. To bridge the gap, Vid2World systematically explores video diffusion causalization, reshaping both the architecture and training objective of pre-trained models to enable autoregressive generation. Additionally, it incorporates a causal action guidance mechanism to enhance action controllability in the resulting interactive world models. Extensive experiments across multiple domains, including robot manipulation, 3D game simulation, and open-world navigation, demonstrate that our method offers a scalable and effective pathway for repurposing highly capable video diffusion models into interactive world models.

[清华]

Generating human grasping poses that accurately reflect both object geometry and user-specified interaction semantics is essential for natural hand-object interactions in AR/VR and embodied AI. However, existing semantic grasping approaches struggle with the large modality gap between 3D object representations and textual instructions, and often lack explicit spatial or semantic constraints, leading to physically invalid or semantically inconsistent grasps. In this work, we present AffordGrasp, a diffusion-based framework that produces physically stable and semantically faithful human grasps with high precision. We first introduce a scalable annotation pipeline that automatically enriches hand-object interaction datasets with fine-grained structured language labels capturing interaction intent. Building upon these annotations, AffordGrasp integrates an affordance-aware latent representation of hand poses with a dual-conditioning diffusion process, enabling the model to jointly reason over object geometry, spatial affordances, and instruction semantics. A distribution adjustment module further enforces physical contact consistency and semantic alignment. We evaluate AffordGrasp across four instruction-augmented benchmarks derived from HO-3D, OakInk, GRAB, and AffordPose, and observe substantial improvements over state-of-the-art methods in grasp quality, semantic accuracy, and diversity.

Understanding spatial affordances -- comprising the contact regions of object interaction and the corresponding contact poses -- is essential for robots to effectively manipulate objects and accomplish diverse tasks. However, existing spatial affordance prediction methods mainly focus on locating the contact regions while delegating the pose to independent pose estimation approaches, which can lead to task failures due to inconsistencies between predicted contact regions and candidate poses. In this work, we propose RoboPCA, a pose-centered affordance prediction framework that jointly predicts task-appropriate contact regions and poses conditioned on instructions. To enable scalable data collection for pose-centered affordance learning, we devise Human2Afford, a data curation pipeline that automatically recovers scene-level 3D information and infers pose-centered affordance annotations from human demonstrations. With Human2Afford, scene depth and the interaction object's mask are extracted to provide 3D context and object localization, while pose-centered affordance annotations are obtained by tracking object points within the contact region and analyzing hand-object interaction patterns to establish a mapping from the 3D hand mesh to the robot end-effector orientation. By integrating geometry-appearance cues through an RGB-D encoder and incorporating mask-enhanced features to emphasize task-relevant object regions into the diffusion-based framework, RoboPCA outperforms baseline methods on image datasets, simulation, and real robots, and exhibits strong generalization across tasks and categories.

Localizing objects and parts from natural language in 3D space is essential for robotics, AR, and embodied AI, yet existing methods face a trade-off between the accuracy and geometric consistency of per-scene optimization and the efficiency of feed-forward inference. We present TrianguLang, a feed-forward framework for 3D localization that requires no camera calibration at inference. Unlike prior methods that treat views independently, we introduce Geometry-Aware Semantic Attention (GASA), which utilizes predicted geometry to gate cross-view feature correspondence, suppressing semantically plausible but geometrically inconsistent matches without requiring ground-truth poses. Validated on five benchmarks including ScanNet++ and uCO3D, TrianguLang achieves state-of-the-art feed-forward text-guided segmentation and localization, reducing user effort from $O(N)$ clicks to a single text query. The model processes each frame at 1008x1008 resolution in $\sim$57ms ($\sim$18 FPS) without optimization, enabling practical deployment for interactive robotics and AR applications. Code and checkpoints are available at https://cwru-aism.github.io/triangulang/.

Establishing dense correspondence across 3D shapes is crucial for fundamental downstream tasks, including texture transfer, shape interpolation, and robotic manipulation. However, learning these mappings without manual supervision remains a formidable challenge, particularly under severe non-isometric deformations and in inter-class settings where geometric cues are ambiguous. Conventional functional map methods, while elegant, typically struggle in these regimes due to their reliance on isometry. To address this, we present GLASS, a framework that bridges the gap by integrating geometric spectral analysis with rich semantic priors from vision-language foundation models. GLASS introduces three key innovations: (i) a view-consistent strategy that enables robust multi-view visual feature extraction from powerful vision foundation models; (ii) the injection of language embeddings into vertex descriptors via zero-shot 3D segmentation, capturing high-level part semantics; and (iii) a graph-assisted contrastive loss that enforces structural consistency between regions (e.g., source's head'' $\leftrightarrow$ target's head'') by leveraging geodesic and topological relationships between regions. This design allows GLASS to learn globally coherent and semantically consistent maps without ground-truth supervision. Extensive experiments demonstrate that GLASS achieves state-of-the-art performance across all regimes, maintaining high accuracy on standard near-isometric tasks while significantly advancing performance in challenging settings. Specifically, it achieves average geodesic errors of 0.21, 4.5, and 5.6 on the inter-class benchmark SNIS and non-isometric benchmarks SMAL and TOPKIDS, reducing errors from URSSM baselines of 0.49, 6.0, and 8.9 by 57%, 25%, and 37%, respectively.

Rigid bodies constitute the smallest manipulable elements in the real world, and understanding how they physically interact is fundamental to embodied reasoning and robotic manipulation. Thus, accurate detection, segmentation, and tracking of moving rigid bodies is essential for enabling reasoning modules to interpret and act in diverse environments. However, current segmentation models trained on semantic grouping are limited in their ability to provide meaningful interaction-level cues for completing embodied tasks. To address this gap, we introduce MotionBit, a novel concept that, unlike prior formulations, defines the smallest unit in motion-based segmentation through kinematic spatial twist equivalence, independent of semantics. In this paper, we contribute (1) the MotionBit concept and definition, (2) a hand-labeled benchmark, called MoRiBo, for evaluating moving rigid-body segmentation across robotic manipulation and human-in-the-wild videos, and (3) a learning-free graph-based MotionBits segmentation method that outperforms state-of-the-art embodied perception methods by 37.3\% in macro-averaged mIoU on the MoRiBo benchmark. Finally, we demonstrate the effectiveness of MotionBits segmentation for downstream embodied reasoning and manipulation tasks, highlighting its importance as a fundamental primitive for understanding physical interactions.

Understanding and localizing objects in complex 3D environments from natural language descriptions, known as 3D Visual Grounding (3DVG), is a foundational challenge in embodied AI, with broad implications for robotics, augmented reality, and human-machine interaction. Large-scale pre-trained foundation models have driven significant progress on this front, enabling open-vocabulary 3DVG that allows systems to locate arbitrary objects in a given scene. However, their reliance on pre-trained models constrains 3D perception and reasoning within the inherited knowledge boundaries, resulting in limited generalization to unseen spatial relationships and poor robustness to out-of-distribution scenes. In this paper, we replace this constrained perception with training-free visual and geometric reasoning, thereby unlocking open-world 3DVG that enables the localization of any object in any scene beyond the training data. Specifically, the proposed UniGround operates in two stages: a Global Candidate Filtering stage that constructs scene candidates through training-free 3D topology and multi-view semantic encoding, and a Local Precision Grounding stage that leverages multi-scale visual prompting and structured reasoning to precisely identify the target object. Experiments on ScanRefer and EmbodiedScan show that UniGround achieves 46.1\%/34.1\% Acc@0.25/0.5 on ScanRefer and 28.7\% Acc@0.25 on EmbodiedScan, establishing a new state-of-the-art among zero-shot methods on EmbodiedScan without any 3D supervision. We further evaluate UniGround in real-world environments under uncontrolled reconstruction conditions and substantial domain shift, showing training-free reasoning generalizes robustly beyond curated benchmarks.