arXiv subjects

Zhe Qian

Publications and source records attributed to Zhe Qian.

11 recordsLinked to original sources

arXiv2026

OC-VLA++: Monocular Geometry-Guided Cross-View Consistency for Viewpoint-Robust Robotic Manipulation

We propose OC-VLA++, an extension of OC-VLA for viewpoint generalization under limited camera coverage. While OC-VLA grounds robot actions in the camera coordinate system to align action supervision with visual observations, camera-space grounding alone can still overfit to the few viewpoints observed during training. OC-VLA++ addresses this limitation by introducing geometry-guided paired-view supervision and an explicit cross-view action-equivariance objective. Given paired observations of the same manipulation scene from geometrically related viewpoints, the model is trained such that their camera-space predictions correspond to the same robot-frame action. This objective explicitly supervises how action predictions should transform across viewpoints, rather than relying solely on image-level augmentation. Experiments demonstrate substantial improvements in unseen-view generalization under limited camera coverage, with performance degrading more gracefully under increasing camera displacement. These results establish cross-view action equivariance as an effective complement to observation-centric action grounding for robust real-world deployment.

cs.RO↗

arXiv2026

Reasoning emerges from constrained inference manifolds in large language models

Reasoning in large language models is predominantly evaluated through labeled benchmarks, conflating task performance with the quality of internal inference. Here we study reasoning as an intrinsic dynamical process by examining the evolution of internal representations during inference. We find that inference-time dynamics consistently self-organize into low-dimensional manifolds embedded within high-dimensional representation spaces. we find that such geometric compression, although pervasive, is not sufficient for stable or reliable reasoning. Instead, effective reasoning dynamics emerge within a constrained structural regime characterized by three conditions: adequate representational expressivity, spontaneous manifold compression, and preservation of non-degenerate information volume within the compressed subspace. Models outside this regime exhibit characteristic pathological inference dynamics. Based on these insights, we introduce a unified, label-free diagnostic computed solely from internal dynamics. These findings suggest that reasoning in LLMs is fundamentally governed by geometric and informational constraints, offering a complementary framework to benchmark-centric assessment.

cs.LG↗

arXiv2026

Cognitive Pivot Points and Visual Anchoring: Unveiling and Rectifying Hallucinations in Multimodal Reasoning Models

Multimodal Large Reasoning Models (MLRMs) have achieved remarkable strides in visual reasoning through test time compute scaling, yet long chain reasoning remains prone to hallucinations. We identify a concerning phenomenon termed the Reasoning Vision Truth Disconnect (RVTD): hallucinations are strongly correlated with cognitive bifurcation points that often exhibit high entropy states. We attribute this vulnerability to a breakdown in visual semantic anchoring, localized within the network's intermediate layers; specifically, during these high uncertainty transitions, the model fails to query visual evidence, reverting instead to language priors. Consequently, we advocate a shift from solely outcome level supervision to augmenting it with fine grained internal attention guidance. To this end, we propose V-STAR (Visual Structural Training with Attention Reinforcement), a lightweight, holistic training paradigm designed to internalize visually aware reasoning capabilities. Central to our approach is the Hierarchical Visual Attention Reward (HVAR), integrated within the GRPO framework. Upon detecting high entropy states, this mechanism dynamically incentivizes visual attention across critical intermediate layers, thereby anchoring the reasoning process back to the visual input. Furthermore, we introduce the Forced Reflection Mechanism (FRM), a trajectory editing strategy that disrupts cognitive inertia by triggering reflection around high entropy cognitive bifurcation points and encouraging verification of subsequent steps against the visual input, thereby translating external debiasing interventions into an intrinsic capability for hallucination mitigation.

cs.AI↗

arXiv2026

SVSR: A Self-Verification and Self-Rectification Paradigm for Multimodal Reasoning

Current multimodal models often suffer from shallow reasoning, leading to errors caused by incomplete or inconsistent thought processes. To address this limitation, we propose Self-Verification and Self-Rectification (SVSR), a unified framework that explicitly integrates self-verification and self-rectification into the model's reasoning pipeline, substantially improving robustness and reliability in complex visual understanding and multimodal reasoning tasks. SVSR is built on a novel three-stage training paradigm. First, we construct a high-quality unified preference dataset by refining reasoning traces from pre-trained vision-language models, incorporating both forward and backward reasoning to embed self-reflective signals. Second, we perform cold-start supervised fine-tuning on this dataset to learn structured, multi-step reasoning behaviors. Third, we apply a Semi-online Direct Preference Optimization (Semi-online DPO) process, continuously augmenting the training corpus with high-quality, model-generated reasoning traces filtered by a powerful teacher VLM. This pipeline enables the model to learn, elicit, and refine its ability to self-verify and self-rectify. Extensive experiments across diverse benchmarks demonstrate that SVSR improves reasoning accuracy and enables stronger generalization to unseen tasks and question types. Notably, once trained with explicit self-reflective reasoning, the model also exhibits improved implicit reasoning ability, outperforming strong baselines even when no explicit reasoning traces are provided. These results highlight the potential of SVSR for building more dependable, introspective, and cognitively aligned multimodal systems.

cs.AI↗

arXiv2026

Region-Constrained Group Relative Policy Optimization for Flow-Based Image Editing

Instruction-guided image editing requires balancing target modification with non-target preservation. Recently, flow-based models have emerged as a strong and increasingly adopted backbone for instruction-guided image editing, thanks to their high fidelity and efficient deterministic ODE sampling. Building on this foundation, GRPO-based reward-driven post-training has been explored to directly optimize editing-specific rewards, improving instruction following and editing consistency. However, existing methods often suffer from noisy credit assignment: global exploration also perturbs non-target regions, inflating within-group reward variance and yielding noisy GRPO advantages. To address this, we propose RC-GRPO-Editing, a region-constrained GRPO post-training framework for flow-based image editing under deterministic ODE sampling. It suppresses background-induced nuisance variance to enable cleaner localized credit assignment, improving editing region instruction adherence while preserving non-target content. Concretely, we localize exploration via region-decoupled initial noise perturbations to reduce background-induced reward variance and stabilize GRPO advantages, and introduce an attention concentration reward that aligns cross-attention with the intended editing region throughout the rollout, reducing unintended changes in non-target regions. Experiments on CompBench show consistent improvements in editing region instruction adherence and non-target preservation.

cs.CV↗

arXiv2026

Thinking in Uncertainty: Mitigating Hallucinations in MLRMs with Latent Entropy-Aware Decoding

Recent advancements in multimodal large reasoning models (MLRMs) have significantly improved performance in visual question answering. However, we observe that transition words (e.g., because, however, and wait) are closely associated with hallucinations and tend to exhibit high-entropy states. We argue that adequate contextual reasoning information can be directly extracted from the token probability distribution. Inspired by superposed representation theory, we propose leveraging latent superposed reasoning to integrate multiple candidate semantics and maintain latent reasoning trajectories. The hypothesis is that reliance on discrete textual inputs may drive the model toward sequential explicit reasoning, underutilizing dense contextual cues during high-entropy reasoning stages. Therefore, we propose constructing rich semantic representations from the token probability distributions to enhance in-context reasoning. With this goal, we present Latent Entropy-Aware Decoding (LEAD), an efficient plug-and-play decoding strategy that leverages semantic context to achieve reliable reasoning. The heart of our method lies in entropy-aware reasoning mode switching. The model employs probability-weighted continuous embeddings under high-entropy states and transitions back to discrete token embeddings as entropy decreases. Moreover, we propose a prior-guided visual anchor injection strategy that encourages the model to focus on visual information. Extensive experiments show that LEAD effectively mitigates hallucinations across various MLRMs on multiple benchmarks.

cs.CV↗

arXiv2025

Geometric Prior-Guided Federated Prompt Calibration

Federated Prompt Learning (FPL) offers a parameter-efficient solution for collaboratively training large models, but its performance is severely hindered by data heterogeneity, which causes locally trained prompts to become biased. Existing methods, focusing on aggregation or regularization, fail to address this root cause of local training bias. To this end, we propose Geometry-Guided Text Prompt Calibration (GGTPC), a novel framework that directly corrects this bias by providing clients with a global geometric prior. This prior, representing the shape of the global data distribution derived from the covariance matrix, is reconstructed on the server in a privacy-preserving manner. Clients then use a novel Geometry-Prior Calibration Layer (GPCL) to align their local feature distributions with this global prior during training. Extensive experiments show GGTPC's effectiveness. On the label-skewed CIFAR-100 dataset ($\beta$=0.1), it outperforms the state-of-the-art by 2.15\%. Under extreme skew ($\beta$=0.01), it improves upon the baseline by 9.17\%. Furthermore, as a plug-and-play module on the domain-skewed Office-Home dataset, it boosts FedAvg's performance by 4.60\%. These results demonstrate that GGTPC effectively mitigates data heterogeneity by correcting the fundamental local training bias, serving as a versatile module to enhance various FL algorithms.

cs.LG↗

arXiv2025

From Perception to Cognition: A Survey of Vision-Language Interactive Reasoning in Multimodal Large Language Models

Multimodal Large Language Models (MLLMs) strive to achieve a profound, human-like understanding of and interaction with the physical world, but often exhibit a shallow and incoherent integration when acquiring information (Perception) and conducting reasoning (Cognition). This disconnect leads to a spectrum of reasoning failures, with hallucination being the most prominent. Collectively, these issues expose a fundamental challenge: the ability to process pixels does not yet confer the ability to construct a coherent, credible internal world model. To systematically dissect and address this challenge, this survey introduces a novel and unified analytical framework: ``From Perception to Cognition." We deconstruct the complex process of vision-language interactive understanding into two interdependent layers: Perception, the foundational ability to accurately extract visual information and achieve fine-grained alignment with textual instructions; and Cognition, the higher-order capability for proactive, multi-step, goal-oriented reasoning built upon this perceptual foundation, the core of which is the formation of a dynamic observe-think-verify reasoning loop. Guided by this framework, this paper systematically analyzes the key bottlenecks of current MLLMs at both layers. It surveys the landscape of cutting-edge methods designed to address these challenges, spanning from techniques that enhance low-level visual representations to those that improve high-level reasoning paradigms. Furthermore, we review critical benchmarks and delineate future research directions. This survey aims to provide the research community with a clear, structured perspective for understanding the intrinsic limitations of current MLLMs and to illuminate the path toward building next-generation models capable of deep reasoning and a genuine understanding of the world.

cs.AI↗

arXiv2024

A Memory Reduction Compact Gas Kinetic Scheme on 3D Unstructured Meshes

This paper introduces a memory-reduction third-order compact gas-kinetic scheme (CGKS) for solving compressible Euler and Navier-Stokes equations on 3D unstructured meshes. The scheme utilizes a time-evolution gas distribution function to provide a time-evolution solution at cell interfaces, enabling the implementation of Hermite WENO techniques for high-order reconstruction. However, the HWENO method needs to store a coefficients matrix for the quadratic polynomial to achieve third-order accuracy, resulting in high memory usage. A novel reconstruction method, built upon HWENO reconstruction, has been designed to enhance computational efficiency and reduce memory usage compared to the original CGKS. The simple idea is that the first-order and second-order terms of the quadratic polynomials are determined in a two-step way. In the first step, the second-order terms are obtained from the reconstruction of a linear polynomial of the first-order derivatives by only using the cell-averaged slopes, since the second-order derivatives are nothing but the "derivatives of derivatives". Subsequently, the first-order terms left can be determined by the linear reconstruction only using cell-averaged values. Thus, we successfully split one quadratic least-square regression into several linear least-square regressions, which are commonly used in a second-order finite volume code. Since only a small matrix inversion is needed in a 3-D linear least-square regression, the computational cost for the new reconstruction is dramatically reduced and the storage of the reconstruction-coefficient matrix is no longer necessary. The proposed new reconstruction technique can reduce the overall computational cost by about 20 to 30 percent. The challenging large-scale unsteady numerical simulation is performed, which demonstrates that the current improvement brings the CGKS to a new level for industrial applications.

math.NA↗

arXiv2017

NaCl crystal from salt solution with far below saturated concentration under ambient condition

Under ambient conditions, we directly observed NaCl crystals experimentally in the rGO membranes soaked in the salt solution with concentration below and far below the saturated concentration. Moreover, in most probability, the NaCl crystals show stoichiometries behavior. We attribute this unexpected crystallization to the cation-π interactions between the ions and the aromatic rings of the rGO.

physics.chem-ph↗

arXiv2016

Highly Efficient Ion Rejection by Graphene Oxide Membranes via Ion-controlling Interlayer Spacing

Because they may provide ultrathin, high-flux, and energy-efficient membranes for precise ionic and molecular sieving in aqueous solution, GO membranes (partially oxidized, stacked sheets of graphene) have shown great potential in water desalination and purification, gas and ion separation, biosensors, proton conductors, lithium-based batteries and super-capacitors. Unlike carbon nanotube (CNT) membranes, in which the nanotube pores have fixed sizes, the pores of GO membranes - the interlayer spacing between GO sheets - are of variable size. This presents a challenge for using GO membranes for filtration. Despite the great efforts to tune and fix the interlayer spacing, it remains difficult both to reduce the interlayer spacing sufficiently to exclude small ions while keeping this separation constant against the tendency of GO membranes to swell when immersed in aqueous solution, which greatly affects the applications of GO membranes. Here, we demonstrate experimentally that highly efficient and selective ion rejection by GO membranes can be readily achieved by controlling the interlayer spacing of GO membranes using cations (K+, Na+, Ca2+, Li+ and Mg2+) themselves. The interspacing can be controlled with precision as small as 1 A, and GO membranes controlled by one kind of cation can exclude other cations with a larger hydrated volume, which can only be accommodated with a larger interlayer spacing. First-principles calculations reveal that the strong noncovalent cation-pi interactions between hydrated cations in solution and aromatic ring structures in GO are the cause of this unexpected behavior. These findings open up new avenues for using GO membranes for water desalination and purification, lithium-based batteries and super-capacitors, molecular sieves for separating ions or molecules, and many other applications.

cond-mat.mtrl-sci↗