ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models

Vision-Language-Action (VLA) models are a powerful paradigm for generalist robotic control. However, their high computational cost and limited control frequency hinder real-time robotic manipulation, especially when large vision-language backbones and iterative action heads run at every control step. Existing VLA acceleration methods often optimize individual components or rely on fixed acceleration rules, treating different control steps with largely fixed computation and overlooking the non-uniform reasoning demands of sequential embodied control. Inspired by human motor control, where cognitive and feedback resources concentrate on goal-sensitive stages, we argue that VLA models should learn when to invest full computation and when to reuse prior computation. We propose ElegantVLA, a plug-in phase-adaptive inference framework that accelerates VLA models through intra-model dynamic compute scheduling. ElegantVLA introduces a lightweight scheduler that observes temporal representation similarity, robot-motion cues, and episode progress to jointly allocate computation across the vision encoder, LLM, and action head. For perception-language reasoning, the scheduler selects a five-level Vision-LLM compute mode, from full recomputation to multi-step temporal reuse, based on visual-language representation stability. For action generation, it selects a three-level denoising mode, reusing intermediate denoising states during stable motion while preserving full refinement for goal-sensitive stages. By coordinating these decisions, ElegantVLA offers a general acceleration framework for modern VLA pipelines with explicit action-generation modules, without modifying or retraining the base model. Experiments on GR00T and CogACT achieve up to 2.55x and 3.77x speedup, and on six real-world GR00T tasks ElegantVLA cuts computation by 2.18x while raising control frequency from 13.8 Hz to 26.3 Hz.

ReCA: Multi-Shot Long Video Extrapolation via Recursive Context Allocation

Minute-scale cinematic video generation is a central challenge for generative video models. Existing paradigms address only fragments of this challenge: single-shot extrapolation preserves an anchor but lacks cinematic structure, while multi-shot storytelling imposes structure yet remains free to invent its visual states rather than continue an observed one. We define Multi-Shot Video Extrapolation (MSVE), a task that extends an observed frame or clip into a sequence of cinematically structured shots while preserving anchor state and advancing narrative intent. This setting operates under the finite per-call generation budget of short-video models. We identify three coupled bottlenecks: (1) global planners over-specify unsupported details from full screenplays; (2) shot-level prompts dilute task-relevant state when carrying the complete story; and (3) temporal chaining turns generated frames into a lossy memory in which identity, scene, object, and action state decay. MSVE reveals that long-video failure is not merely a limitation of context length, but a failure of context allocation. We propose Recursive Context Allocation (ReCA), an inference-time framework that allocates context hierarchically across planning and generation. ReCA recursively decomposes MSVE into context-bounded subproblems, invokes frozen generators at leaf nodes, and propagates structured state updates across time. To evaluate this setting, we further propose MSVE-Bench and NB-Q, a source-grounded protocol with prompts purpose-built for 3 to 5 minute long-video generation, a regime not addressed by existing short-clip benchmarks. Compared to previous methods, ReCA improves average normalized score by 8 to 16 percent over the strongest competing controller and improves multi-shot consistency metrics by 28 to 43 percent. View the project page at https://reca.vmv.re.

GE-Sim 2.0: A Roadmap Towards Comprehensive Closed-loop Video World Simulators for Robotic Manipulation

We introduce GE-Sim 2.0 (Genie Envisioner World Simulator 2.0), a closed-loop video world simulator for robotic manipulation. Building on the action-conditioned video generation framework of Genie Envisioner, GE-Sim 2.0 is re-trained on thousands of hours of real-world robot data spanning teleoperation, contact-rich interaction, and on-robot policy deployment, substantially improving action-following fidelity and trajectory coverage. On top of this foundation, three new modules close the loop from video simulation to policy learning: a state expert that decodes proprioceptive state from video latents to support next-chunk prediction by downstream VLA policies; a world judge that scores generated rollouts against task instructions, yielding machine-verifiable success signals and rewards in place of manual inspection; and an acceleration framework that delivers a 25-frame rollout in 2.3 seconds on a single H100, with up to 4* frame skipping at inference for long-horizon evaluation. GE-Sim 2.0 tops the public WorldArena leaderboard at only 2B parameters, outperforming both dedicated robotic world models and closed-source general video generators, and policies trained against its rollouts and rewards translate into measurable real-world gains, establishing GE-Sim 2.0 as a practical platform for scalable evaluation and closed-loop learning of manipulation policies.

Test-time Sparsity for Extreme Fast Action Diffusion

Action diffusion excels at high-fidelity action generation but incurs heavy computational costs owing to its iterative denoising nature. Despite current technologies showing promise in accelerating diffusion transformers by reusing the cached features, they struggle to adapt to policy dynamics arising from diverse perceptions and multi-round rollout iterations in open environments. We propose test-time sparsity to tackle this challenge, which aims to accelerate action diffusion by dynamically predicting prunable residual computations for each model forward at test time. However, two bottlenecks remain in this paradigm: 1) repetitive conditional encoding and pruning offset most potential speed gains, and 2) the features cached from previous denoising timesteps cannot constrain large pruning errors under aggressive sparsity. To address the first bottleneck, we design a highly parallelized inference pipeline that minimizes the non-decoder delay to milliseconds. Specifically, we first design a lightweight pruner that shares the encoder with the diffusion transformer. Then, we decouple the encoding and pruning from the autoregressive denoising loop by processing all denoising timesteps in parallel, and overlap the pruner with the decoder forward inference through asynchronism. To overcome the second bottleneck, we introduce an omnidirectional reusing strategy, which achieves 95% sparsity by selectively reusing features cached from the current forward, previous denoising timesteps, and earlier rollout iterations. To learn the rollout-level reusing strategies, we sample a few action trajectories to supervise the sparsified diffusion step by step. Extensive experiments demonstrate that our method reduces FLOPs by 92% and accelerates action generation by 5x, achieving lossless performance with an inference frequency of 47.5 Hz. Our code is available at https://github.com/ky-ji/Test-time-Sparsity.

Unlocking Optical Prior: Spectrum-Guided Knowledge Transfer for SAR Generalized Category Discovery

Generalized Category Discovery (GCD) holds significant promise for the label-scarce Synthetic Aperture Radar (SAR) domain, yet its efficacy is severely constrained by the cross-modal incompatibility between the inherent optical prior of the Large Vision Models (LVMs) and SAR imagery. Existing domain adaptation methods often lack an inductive bias that reflects imaging characteristics, consequently failing to effectively transfer optical prior into the SAR domain. To address this issue, the Modal Discrepancy Curve (MDC) is introduced to model cross-modal discrepancy as a structured frequency-domain descriptor derived from spectral energy distributions. Leveraging this formulation, we propose the MDC-guided Cross-modal Prior Transfer (MCPT) framework, a pre-training paradigm that operates on paired optical-SAR data. Within this framework, Adaptive Frequency Tokenization (AFT) converts the MDC into learnable tokens, and Frequency-aware Expert Refinement (FER) performs band-wise discrepancy-aware feature refinement using these tokens. Based on the refined representations, contrastive learning aligns refined embeddings across modalities and internalizes the adaptation pattern. Ultimately, the superior SAR feature representation capability learned during paired pre-training is applied to downstream single-modal SAR-GCD tasks. Extensive experiments demonstrate state-of-the-art performance across multiple mainstream datasets, indicating that frequency-domain discrepancy modeling enables more effective adaptation of optical prior to SAR imagery.

IP-SAM: Prompt-Space Conditioning for Prompt-Absent Camouflaged Object Detection

Prompt-conditioned foundation segmenters have emerged as a dominant paradigm for image segmentation, where explicit spatial prompts (e.g., points, boxes, masks) guide mask decoding. However, many real-world deployments require fully automatic segmentation, creating a structural mismatch: the decoder expects prompts that are unavailable at inference. Existing adaptations typically modify intermediate features, inadvertently bypassing the model's native prompt interface and weakening prompt-conditioned decoding. We propose IP-SAM, which revisits adaptation from a prompt-space perspective through prompt-space conditioning. Specifically, a Self-Prompt Generator (SPG) distills image context into complementary intrinsic prompts that serve as coarse regional anchors. These cues are projected through SAM2's frozen prompt encoder, restoring prompt-guided decoding without external intervention. To suppress background-induced false positives, Prompt-Space Gating (PSG) leverages the intrinsic background prompt as an asymmetric suppressive constraint prior to decoding. Under a deterministic no-external-prompt protocol, IP-SAM achieves state-of-the-art performance across four camouflaged object detection benchmarks (e.g., MAE 0.017 on COD10K) with only 21.26M trainable parameters (optimizing SPG, PSG, and a task-specific mask decoder trained from scratch, alongside image-encoder LoRA while keeping the prompt encoder frozen). Furthermore, the proposed conditioning strategy generalizes beyond COD to medical polyp segmentation, where a model trained solely on Kvasir-SEG exhibits strong zero-shot transfer to both CVC-ClinicDB and ETIS.

Set-Valued Prediction for Large Language Models with Feasibility-Aware Coverage Guarantees

Large language models (LLMs) inherently operate over a large generation space, yet conventional usage typically reports the most likely generation (MLG) as a point prediction, which underestimates the model's capability: although the top-ranked response can be incorrect, valid answers may still exist within the broader output space and can potentially be discovered through repeated sampling. This observation motivates moving from point prediction to set-valued prediction, where the model produces a set of candidate responses rather than a single MLG. In this paper, we propose a principled framework for set-valued prediction, which provides feasibility-aware coverage guarantees. We show that, given the finite-sampling nature of LLM generation, coverage is not always achievable: even with multiple samplings, LLMs may fail to yield an acceptable response for certain questions within the sampled candidate set. To address this, we establish a minimum achievable risk level (MRL), below which statistical coverage guarantees cannot be satisfied. Building on this insight, we then develop a data-driven calibration procedure that constructs prediction sets from sampled responses by estimating a rigorous threshold, ensuring that the resulting set contains a correct answer with a desired probability whenever the target risk level is feasible. Extensive experiments on six language generation tasks with five LLMs demonstrate both the statistical validity and the predictive efficiency of our framework.

Variable-Resolution Virtual Maps for Autonomous Exploration with Unmanned Surface Vehicles (USVs)

Autonomous exploration by unmanned surface vehicles (USVs) in near-shore waters requires reliable localisation and consistent mapping over extended areas, but this is challenged by GNSS degradation, environment-induced localisation uncertainty, and limited on-board computation. Virtual map-based methods explicitly model localisation and mapping uncertainty by tightly coupling factor-graph SLAM with a map uncertainty criterion. However, their storage and computational costs scale poorly with fixed-resolution workspace discretisations, leading to inefficiency in large near-shore environments. Moreover, overvaluing feature-sparse open-water regions can increase the risk of SLAM failure as a result of imbalance between exploration and exploitation. To address these limitations, we propose a Variable-Resolution Virtual Map (VRVM), a computationally efficient method for representing map uncertainty using bivariate Gaussian virtual landmarks placed in the cells of an adaptive quadtree. The adaptive quadtree enables an area-weighted uncertainty representation that keeps coarse, far-field virtual landmarks deliberately uncertain while allocating higher resolution to information-dense regions, and reduces the sensitivity of the map valuation to local refinements of the tree. An expectation-maximisation (EM) planner is adopted to evaluate pose and map uncertainty along frontiers using the VRVM, balancing exploration and exploitation. We evaluate VRVM against several state-of-the-art exploration algorithms in the VRX Gazebo simulator, using a realistic marina environment across different testing scenarios with an increasing level of exploration difficulty. The results indicate that our method offers safer behaviour and better utilisation of on-board computation in GNSS-degraded near-shore environments.

RoboStream: Weaving Spatio-Temporal Reasoning with Memory in Vision-Language Models for Robotics

Enabling reliable long-horizon robotic manipulation is a crucial step toward open-world embodied intelligence. However, VLM-based planners treat each step as an isolated observation-to-action mapping, forcing them to reinfer scene geometry from raw pixels at every decision point while remaining unaware of how prior actions have reshaped the environment. Despite strong short-horizon performance, these systems lack the spatio-temporal reasoning required for persistent geometric anchoring and memory of action-triggered state transitions. Without persistent state tracking, perceptual errors accumulate across the execution horizon, temporarily occluded objects are catastrophically forgotten, and these compounding failures lead to precondition violations that cascade through subsequent steps. In contrast, humans maintain a persistent mental model that continuously tracks spatial relations and action consequences across interactions rather than reconstructing them at each instant. Inspired by this human capacity for causal spatio-temporal reasoning with persistent memory, we propose RoboStream, a training-free framework that achieves geometric anchoring through Spatio-Temporal Fusion Tokens (STF-Tokens), which bind visual evidence to 3D geometric attributes for persistent object grounding, and maintains causal continuity via a Causal Spatio-Temporal Graph (CSTG) that records action-triggered state transitions across steps. This design enables the planner to trace causal chains and preserve object permanence under occlusion without additional training or fine-tuning. RoboStream achieves 90.5% on long-horizon RLBench and 44.4% on challenging real-world block-building tasks, where both SoFar and VoxPoser score 11.1%, demonstrating that spatio-temporal reasoning and causal memory are critical missing components for reliable long-horizon manipulation.

Sparse ActionGen: Accelerating Diffusion Policy with Real-time Pruning

Diffusion Policy has dominated action generation due to its strong capabilities for modeling multi-modal action distributions, but its multi-step denoising processes make it impractical for real-time visuomotor control. Existing caching-based acceleration methods typically rely on $\textit{static}$ schedules that fail to adapt to the $\textit{dynamics}$ of robot-environment interactions, thereby leading to suboptimal performance. In this paper, we propose $\underline{\textbf{S}}$parse $\underline{\textbf{A}}$ction$\underline{\textbf{G}}$en ($\textbf{SAG}$) for extremely sparse action generation. To accommodate the iterative interactions, SAG customizes a rollout-adaptive prune-then-reuse mechanism that first identifies prunable computations globally and then reuses cached activations to substitute them during action diffusion. To capture the rollout dynamics, SAG parameterizes an observation-conditioned diffusion pruner for environment-aware adaptation and instantiates it with a highly parameter- and inference-efficient design for real-time prediction. Furthermore, SAG introduces a one-for-all reusing strategy that reuses activations across both timesteps and blocks in a zig-zag manner, minimizing the global redundancy. Extensive experiments on multiple robotic benchmarks demonstrate that SAG achieves up to 4$\times$ generation speedup without sacrificing performance. Project Page: https://sparse-actiongen.github.io/.