arXiv subjects

Yucheng Zhao

Publications and source records attributed to Yucheng Zhao.

At least 19 recordsLinked to original sources

arXiv2026

BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI Agents

Mobile GUI agents are shifting from multi-module frameworks to native models trained end-to-end, yet industrial deployment faces three persistent gaps. Sandbox training produces a distribution mismatch with production environments; expensive real-device failures remain underutilized; and fixed benchmarks saturate, losing the power to guide iteration. We present BlueLM-GUI, a 35B-A3B mobile GUI agent built as a real-device-centric flywheel that closes these gaps through three principles. Every Sample Matters: a dual-track pipeline with Heterogeneous Triple-System Consensus evaluation and an Error Correction \& Derivation Module salvages every trajectory into usable supervision. Every Rollout Is Real: a three-stage recipe---continual pre-training, supervised fine-tuning, and agentic reinforcement learning on hundreds of real phones---grounds every rollout in real production environments, so the capability the model learns transfers directly to deployment. Every Query Evolves: a quota-driven benchmark methodology with three orthogonal axes enables precise attribution and allows the benchmark to be systematically upgraded as the model improves. BlueLM-GUI achieves 87.4 on MobileGUI-VBench, surpassing the best closed-source model by 5.1 points, and 84.9 on AndroidWorld, the best result among open-source models and competitive with closed-source models. These results demonstrate that grounding model training and iterative improvement in both real devices and the three Every principles yields strong, robust, and transferable mobile GUI capability.

cs.AI↗

arXiv2026

Bridging extrinsic and intrinsic variable importance

Variable importance may describe either intrinsic predictive information in a population or extrinsic importance for a fitted prediction rule. Quantifying the uncertainty in variable importance estimates is critical for interpretation. Methods for estimating intrinsic variable importance (we will refer to these as VIMP) and the minipatch leave-one-covariate-out procedure (MPLOCO) target intrinsic and extrinsic importance, respectively, and provide methods for computing standard errors. These two approaches have a shared structure, comparing prediction performance with and without features, but the relationship between them has not been formally characterized. We establish conditions under which the two perspectives align. Under squared-error loss, if the fitted full and reduced learners converge to their oracle counterparts sufficiently fast, then MPLOCO is asymptotically equivalent to VIMP. We provide further conditions extending this result to general loss functions and formalize grouped MPLOCO for potentially overlapping feature groups. Through simulations, we show that VIMP and MPLOCO agree most closely when the fitted learner is well aligned with the data-generating mechanism. In a high-dimensional grouped simulation, both procedures identified the signal-containing groups. In an analysis of HIV-1 VRC01 neutralization sensitivity, both methods placed the same three biologically relevant feature groups among their highest-ranked groups. These results clarify when intrinsic and extrinsic importance can be interpreted similarly and when they provide complementary information.

stat.ME↗

arXiv2026

Defending Jailbreak Attacks on Large Language Models via Manifold Trajectory Kinetics

Jailbreak prompts can bypass alignment guardrails in large language models (LLMs) and elicit unsafe outputs, making reliable deployment-time detection critical. Prior detection approaches largely rely on a fixed metric space, e.g., raw inputs, gradients, or hidden features, in which benign and jailbreak prompts are linearly separable. We show this assumption breaks under (i) pseudo-malicious prompts that are benign by intent but contain safety-related keywords, and (ii) adaptive attacks that explicitly optimize against the deployed detector. To overcome this limitation, we shift our focus from identifying a universal metric space to analyzing the more robust neighborhood structure of the underlying data manifold. We present Manifold Trajectory Kinetics (MTK), which treats an LLM as a kinetic system transforming inputs into outputs and detects jailbreaks by tracking how a prompt's neighborhood structure evolves across layers. Benign prompts remain close to benign neighborhoods throughout inference, whereas jailbreak prompts exhibit a characteristic trajectory that begins near malicious seeds and later strategically shifts toward benign neighborhoods to evade refusal.Across four LLMs and ten jailbreak attacks, MTK achieves strong robustness to both failure modes: on pseudo-malicious prompts, it attains a jailbreak true positive rate of 95% at a false positive rate of 5% on benign prompts and 2% on pseudo-malicious prompts, and under adaptive attacks, it maintains a true positive rate of 85%. We further demonstrate the superior performance of MTK for jailbreak detection in vision-language models. Our code is available at https://github.com/Rookie143/mtk.

cs.CR↗

arXiv2026

Realtime-VLA FLASH: Speculative Inference Framework for Diffusion-based VLAs

Diffusion-based vision-language-action models (dVLAs) are promising for embodied intelligence but are fundamentally limited in real-time deployment by the high latency of full inference. We propose Realtime-VLA FLASH, a speculative inference framework that eliminates most full inference calls during replanning by introducing a lightweight draft model with parallel verification via the main model's Action Expert and a phase-aware fallback mechanism that reverts to the full inference pipeline when needed. This design enables low-latency, high-frequency replanning without sacrificing reliability. Experiments show that on LIBERO, FLASH largely preserves task performance by replacing many 58.0 ms full-inference rounds with speculative rounds as fast as 7.8 ms, lowering task-level average inference latency to 19.1 ms (3.04x speedup). We additionally demonstrate effectiveness on real-world conveyor-belt sorting, highlighting its practical impact for latency-critical embodied tasks.

cs.RO↗

arXiv2026

Robotic Scene Cloning:Advancing Zero-Shot Robotic Scene Adaptation in Manipulation via Visual Prompt Editing

Modern robots can perform a wide range of simple tasks and adapt to diverse scenarios in the well-trained environment. However, deploying pre-trained robot models in real-world user scenarios remains challenging due to their limited zero-shot capabilities, often necessitating extensive on-site data collection. To address this issue, we propose Robotic Scene Cloning (RSC), a novel method designed for scene-specific adaptation by editing existing robot operation trajectories. RSC achieves accurate and scene-consistent sample generation by leveraging a visual prompting mechanism and a carefully tuned condition injection module. Not only transferring textures but also performing moderate shape adaptations in response to the visual prompts, RSC demonstrates reliable task performance across a variety of object types. Experiments across various simulated and real-world environments demonstrate that RSC significantly enhances policy generalization in target environments.

cs.RO↗

arXiv2026

DM0: An Embodied-Native Vision-Language-Action Model towards Physical AI

Moving beyond the traditional paradigm of adapting internet-pretrained models to physical tasks, we present DM0, an Embodied-Native Vision-Language-Action (VLA) framework designed for Physical AI. Unlike approaches that treat physical grounding as a fine-tuning afterthought, DM0 unifies embodied manipulation and navigation by learning from heterogeneous data sources from the onset. Our methodology follows a comprehensive three-stage pipeline: Pretraining, Mid-Training, and Post-Training. First, we conduct large-scale unified pretraining on the Vision-Language Model (VLM) using diverse corpora--seamlessly integrating web text, autonomous driving scenarios, and embodied interaction logs-to jointly acquire semantic knowledge and physical priors. Subsequently, we build a flow-matching action expert atop the VLM. To reconcile high-level reasoning with low-level control, DM0 employs a hybrid training strategy: for embodied data, gradients from the action expert are not backpropagated to the VLM to preserve generalized representations, while the VLM remains trainable on non-embodied data. Furthermore, we introduce an Embodied Spatial Scaffolding strategy to construct spatial Chain-of-Thought (CoT) reasoning, effectively constraining the action solution space. Experiments on the RoboChallenge benchmark demonstrate that DM0 achieves state-of-the-art performance in both Specialist and Generalist settings on Table30.

cs.RO↗

arXiv2025

Dexbotic: Open-Source Vision-Language-Action Toolbox

In this paper, we present Dexbotic, an open-source Vision-Language-Action (VLA) model toolbox based on PyTorch. It aims to provide a one-stop VLA research service for professionals in the field of embodied intelligence. It offers a codebase that supports multiple mainstream VLA policies simultaneously, allowing users to reproduce various VLA methods with just a single environment setup. The toolbox is experiment-centric, where the users can quickly develop new VLA experiments by simply modifying the Exp script. Moreover, we provide much stronger pretrained models to achieve great performance improvements for state-of-the-art VLA policies. Dexbotic will continuously update to include more of the latest pre-trained foundation models and cutting-edge VLA models in the industry.

cs.RO↗

arXiv2025

RoboChallenge: Large-scale Real-robot Evaluation of Embodied Policies

Testing on real machines is indispensable for robotic control algorithms. In the context of learning-based algorithms, especially VLA models, demand for large-scale evaluation, i.e. testing a large number of models on a large number of tasks, is becoming increasingly urgent. However, doing this right is highly non-trivial, especially when scalability and reproducibility is taken into account. In this report, we describe our methodology for constructing RoboChallenge, an online evaluation system to test robotic control algorithms, and our survey of recent state-of-the-art VLA models using our initial benchmark Table30.

cs.RO↗

arXiv2025

ManiAgent: An Agentic Framework for General Robotic Manipulation

While Vision-Language-Action (VLA) models have demonstrated impressive capabilities in robotic manipulation, their performance in complex reasoning and long-horizon task planning is limited by data scarcity and model capacity. To address this, we introduce ManiAgent, an agentic architecture for general manipulation tasks that achieves end-to-end output from task descriptions and environmental inputs to robotic manipulation actions. In this framework, multiple agents involve inter-agent communication to perform environmental perception, sub-task decomposition and action generation, enabling efficient handling of complex manipulation scenarios. Evaluations show ManiAgent achieves an 86.8% success rate on the SimplerEnv benchmark and 95.8% on real-world pick-and-place tasks, enabling efficient data collection that yields VLA models with performance comparable to those trained on human-annotated datasets. The project webpage is available at https://yi-yang929.github.io/ManiAgent/.

cs.RO↗

arXiv2025

IntentionVLA: Generalizable and Efficient Embodied Intention Reasoning for Human-Robot Interaction

Vision-Language-Action (VLA) models leverage pretrained vision-language models (VLMs) to couple perception with robotic control, offering a promising path toward general-purpose embodied intelligence. However, current SOTA VLAs are primarily pretrained on multimodal tasks with limited relevance to embodied scenarios, and then finetuned to map explicit instructions to actions. Consequently, due to the lack of reasoning-intensive pretraining and reasoning-guided manipulation, these models are unable to perform implicit human intention reasoning required for complex, real-world interactions. To overcome these limitations, we propose \textbf{IntentionVLA}, a VLA framework with a curriculum training paradigm and an efficient inference mechanism. Our proposed method first leverages carefully designed reasoning data that combine intention inference, spatial grounding, and compact embodied reasoning, endowing the model with both reasoning and perception capabilities. In the following finetuning stage, IntentionVLA employs the compact reasoning outputs as contextual guidance for action generation, enabling fast inference under indirect instructions. Experimental results show that IntentionVLA substantially outperforms $π_0$, achieving 18\% higher success rates with direct instructions and 28\% higher than ECoT under intention instructions. On out-of-distribution intention tasks, IntentionVLA achieves over twice the success rate of all baselines, and further enables zero-shot human-robot interaction with 40\% success rate. These results highlight IntentionVLA as a promising paradigm for next-generation human-robot interaction (HRI) systems.

cs.RO↗

arXiv2025

LLaDA-VLA: Vision Language Diffusion Action Models

The rapid progress of auto-regressive vision-language models (VLMs) has inspired growing interest in vision-language-action models (VLA) for robotic manipulation. Recently, masked diffusion models, a paradigm distinct from autoregressive models, have begun to demonstrate competitive performance in text generation and multimodal applications, leading to the development of a series of diffusion-based VLMs (d-VLMs). However, leveraging such models for robot policy learning remains largely unexplored. In this work, we present LLaDA-VLA, the first Vision-Language-Diffusion-Action model built upon pretrained d-VLMs for robotic manipulation. To effectively adapt d-VLMs to robotic domain, we introduce two key designs: (1) a localized special-token classification strategy that replaces full-vocabulary classification with special action token classification, reducing adaptation difficulty; (2) a hierarchical action-structured decoding strategy that decodes action sequences hierarchically considering the dependencies within and across actions. Extensive experiments demonstrate that LLaDA-VLA significantly outperforms state-of-the-art VLAs on both simulation and real-world robots.

cs.RO↗

arXiv2025

Hita: Holistic Tokenizer for Autoregressive Image Generation

Vanilla autoregressive image generation models generate visual tokens step-by-step, limiting their ability to capture holistic relationships among token sequences. Moreover, because most visual tokenizers map local image patches into latent tokens, global information is limited. To address this, we introduce \textit{Hita}, a novel image tokenizer for autoregressive (AR) image generation. It introduces a holistic-to-local tokenization scheme with learnable holistic queries and local patch tokens. Hita incorporates two key strategies to better align with the AR generation process: 1) {arranging} a sequential structure with holistic tokens at the beginning, followed by patch-level tokens, and using causal attention to maintain awareness of previous tokens; and 2) adopting a lightweight fusion module before feeding the de-quantized tokens into the decoder to control information flow and prioritize holistic tokens. Extensive experiments show that Hita accelerates the training speed of AR generators and outperforms those trained with vanilla tokenizers, achieving \textbf{2.59 FID} and \textbf{281.9 IS} on the ImageNet benchmark. Detailed analysis of the holistic representation highlights its ability to capture global image properties, such as textures, materials, and shapes. Additionally, Hita also demonstrates effectiveness in zero-shot style transfer and image in-painting. The code is available at \href{https://github.com/CVMI-Lab/Hita}{https://github.com/CVMI-Lab/Hita}.

cs.CV↗

arXiv2025

ROSA: Harnessing Robot States for Vision-Language and Action Alignment

Vision-Language-Action (VLA) models have recently made significant advance in multi-task, end-to-end robotic control, due to the strong generalization capabilities of Vision-Language Models (VLMs). A fundamental challenge in developing such models is effectively aligning the vision-language space with the robotic action space. Existing approaches typically rely on directly fine-tuning VLMs using expert demonstrations. However, this strategy suffers from a spatio-temporal gap, resulting in considerable data inefficiency and heavy reliance on human labor. Spatially, VLMs operate within a high-level semantic space, whereas robotic actions are grounded in low-level 3D physical space; temporally, VLMs primarily interpret the present, while VLA models anticipate future actions. To overcome these challenges, we propose a novel training paradigm, ROSA, which leverages robot state estimation to improve alignment between vision-language and action spaces. By integrating robot state estimation data obtained via an automated process, ROSA enables the VLA model to gain enhanced spatial understanding and self-awareness, thereby boosting performance and generalization. Extensive experiments in both simulated and real-world environments demonstrate the effectiveness of ROSA, particularly in low-data regimes.

cs.RO↗

arXiv2025

PADriver: Towards Personalized Autonomous Driving

In this paper, we propose PADriver, a novel closed-loop framework for personalized autonomous driving (PAD). Built upon Multi-modal Large Language Model (MLLM), PADriver takes streaming frames and personalized textual prompts as inputs. It autoaggressively performs scene understanding, danger level estimation and action decision. The predicted danger level reflects the risk of the potential action and provides an explicit reference for the final action, which corresponds to the preset personalized prompt. Moreover, we construct a closed-loop benchmark named PAD-Highway based on Highway-Env simulator to comprehensively evaluate the decision performance under traffic rules. The dataset contains 250 hours videos with high-quality annotation to facilitate the development of PAD behavior analysis. Experimental results on the constructed benchmark show that PADriver outperforms state-of-the-art approaches on different evaluation metrics, and enables various driving modes.

cs.CV↗

arXiv2025

Simulating Nighttime Visible Satellite Imagery of Tropical Cyclones Using Conditional Generative Adversarial Networks

Visible (VIS) imagery is important for monitoring Tropical Cyclones (TCs) but is unavailable at night. This study presents a Conditional Generative Adversarial Networks (CGAN) model to generate nighttime VIS imagery with significantly enhanced accuracy and spatial resolution. Our method offers three key improvements compared to existing models. First, we replaced the L1 loss in the pix2pix framework with the Structural Similarity Index Measure (SSIM) loss, which significantly reduced image blurriness. Second, we selected multispectral infrared (IR) bands as input based on a thorough examination of their spectral properties, providing essential physical information for accurate simulation. Third, we incorporated the direction parameters of the sun and the satellite, which addressed the dependence of VIS images on sunlight directions and enabled a much larger training set from continuous daytime data. The model was trained and validated using data from the Advanced Himawari Imager (AHI) in the daytime, achieving statistical results of SSIM = 0.923 and Root Mean Square Error (RMSE) = 0.0299, which significantly surpasses existing models. We also performed a cross-satellite nighttime model validation using the Day/Night Band (DNB) of the Visible/Infrared Imager Radiometer Suite (VIIRS), which yields outstanding results compared to existing models. Our model is operationally applied to generate accurate VIS imagery with arbitrary virtual sunlight directions, significantly contributing to the nighttime monitoring of various meteorological phenomena.

physics.ao-ph↗

arXiv2025

Ross3D: Reconstructive Visual Instruction Tuning with 3D-Awareness

The rapid development of Large Multimodal Models (LMMs) for 2D images and videos has spurred efforts to adapt these models for interpreting 3D scenes. However, the absence of large-scale 3D vision-language datasets has posed a significant obstacle. To address this issue, typical approaches focus on injecting 3D awareness into 2D LMMs by designing 3D input-level scene representations. This work provides a new perspective. We introduce reconstructive visual instruction tuning with 3D-awareness (Ross3D), which integrates 3D-aware visual supervision into the training procedure. Specifically, it incorporates cross-view and global-view reconstruction. The former requires reconstructing masked views by aggregating overlapping information from other views. The latter aims to aggregate information from all available views to recover Bird's-Eye-View images, contributing to a comprehensive overview of the entire scene. Empirically, Ross3D achieves state-of-the-art performance across various 3D scene understanding benchmarks. More importantly, our semi-supervised experiments demonstrate significant potential in leveraging large amounts of unlabeled 3D vision-only data.

cs.CV↗

arXiv2025

CSF-Net: Cross-Modal Spatiotemporal Fusion Network for Pulmonary Nodule Malignancy Predicting

Pulmonary nodules are an early sign of lung cancer, and detecting them early is vital for improving patient survival rates. Most current methods use only single Computed Tomography (CT) images to assess nodule malignancy. However, doctors typically make a comprehensive assessment in clinical practice by integrating follow-up CT scans with clinical data. To enhance this process, our study introduces a Cross-Modal Spatiotemporal Fusion Network, named CSF-Net, designed to predict the malignancy of pulmonary nodules using follow-up CT scans. This approach simulates the decision-making process of clinicians who combine follow-up imaging with clinical information. CSF-Net comprises three key components: spatial feature extraction module, temporal residual fusion module, and cross-modal attention fusion module. Together, these modules enable precise predictions of nodule malignancy. Additionally, we utilized the publicly available NLST dataset to screen and annotate the specific locations of pulmonary nodules and created a new dataset named NLST-cmst. Our experimental results on the NLST-cmst dataset demonstrate significant performance improvements, with an accuracy of 0.8974, a precision of 0.8235, an F1 score of 0.8750, an AUC of 0.9389, and a recall of 0.9333. These findings indicate that our multimodal spatiotemporal fusion approach, which combines follow-up data with clinical information, surpasses existing methods, underscoring its effectiveness in predicting nodule malignancy.

eess.IV↗

arXiv2024

Reconstructive Visual Instruction Tuning

This paper introduces reconstructive visual instruction tuning (ROSS), a family of Large Multimodal Models (LMMs) that exploit vision-centric supervision signals. In contrast to conventional visual instruction tuning approaches that exclusively supervise text outputs, ROSS prompts LMMs to supervise visual outputs via reconstructing input images. By doing so, it capitalizes on the inherent richness and detail present within input images themselves, which are often lost in pure text supervision. However, producing meaningful feedback from natural images is challenging due to the heavy spatial redundancy of visual signals. To address this issue, ROSS employs a denoising objective to reconstruct latent representations of input images, avoiding directly regressing exact raw RGB values. This intrinsic activation design inherently encourages LMMs to maintain image detail, thereby enhancing their fine-grained comprehension capabilities and reducing hallucinations. Empirically, ROSS consistently brings significant improvements across different visual encoders and language models. In comparison with extrinsic assistance state-of-the-art alternatives that aggregate multiple visual experts, ROSS delivers competitive performance with a single SigLIP visual encoder, demonstrating the efficacy of our vision-centric supervision tailored for visual outputs.

cs.CV↗