arXiv subjects

Zhanyu Ma

Publications and source records attributed to Zhanyu Ma.

At least 55 records · Page 3Linked to original sources

arXiv2026

DriveRX: A Vision-Language Reasoning Model for Cross-Task Autonomous Driving

Effective autonomous driving hinges on robust reasoning across perception, prediction, planning, and behavior. However, conventional end-to-end models fail to generalize in complex scenarios due to the lack of structured reasoning. While recent vision-language models (VLMs) have been applied to driving tasks, they typically rely on isolated modules and static supervision, limiting their ability to support multi-stage decision-making. We present AutoDriveRL, a unified training framework that formulates autonomous driving as a structured reasoning process over four core tasks. Each task is independently modeled as a vision-language QA problem and optimized using task-specific reward models, enabling fine-grained reinforcement signals at different reasoning stages. Within this framework, we train DriveRX, a cross-task reasoning VLM designed for multi-stage decision-making. DriveRX achieves strong performance on the public benchmark, outperforming GPT-4o in behavior reasoning and demonstrating robustness under complex or corrupted driving conditions. DriveRX serves as a high-level semantic reasoning backbone, producing structured stage-wise reasoning chains that enhance decision consistency. These outputs also provide high-quality supervisory signals for annotation and downstream planning/control models. We release the AutoDriveRL framework and DriveRX to support future research.

cs.CV↗

arXiv2026

Fine-Mem: Fine-Grained Feedback Alignment for Long-Horizon Memory Management

Effective memory management is essential for large language model agents to navigate long-horizon tasks. Recent research has explored using Reinforcement Learning to develop specialized memory manager agents. However, existing approaches rely on final task performance as the primary reward, which results in severe reward sparsity and ineffective credit assignment, providing insufficient guidance for individual memory operations. To this end, we propose Fine-Mem, a unified framework designed for fine-grained feedback alignment. First, we introduce a Chunk-level Step Reward to provide immediate step-level supervision via auxiliary chunk-specific question answering tasks. Second, we devise Evidence-Anchored Reward Attribution to redistribute global rewards by anchoring credit to key memory operations, based on the specific memory items utilized as evidence in reasoning. Together, these components enable stable policy optimization and align local memory operations with the long-term utility of memory. Experiments on Memalpha and MemoryAgentBench demonstrate that Fine-Mem consistently outperforms strong baselines, achieving superior success rates across various sub-tasks. Further analysis reveals its adaptability and strong generalization capabilities across diverse model configurations and backbones.

cs.CL↗

arXiv2026

VisualCloze: A Universal Image Generation Framework via Visual In-Context Learning

Recent progress in diffusion models significantly advances various image generation tasks. However, the current mainstream approach remains focused on building task-specific models, which have limited efficiency when supporting a wide range of different needs. While universal models attempt to address this limitation, they face critical challenges, including generalizable task instruction, appropriate task distributions, and unified architectural design. To tackle these challenges, we propose VisualCloze, a universal image generation framework, which supports a wide range of in-domain tasks, generalization to unseen ones, unseen unification of multiple tasks, and reverse generation. Unlike existing methods that rely on language-based task instruction, leading to task ambiguity and weak generalization, we integrate visual in-context learning, allowing models to identify tasks from visual demonstrations. Meanwhile, the inherent sparsity of visual task distributions hampers the learning of transferable knowledge across tasks. To this end, we introduce Graph200K, a graph-structured dataset that establishes various interrelated tasks, enhancing task density and transferable knowledge. Furthermore, we uncover that our unified image generation formulation shared a consistent objective with image infilling, enabling us to leverage the strong generative priors of pre-trained infilling models without modifying the architectures.

cs.CV↗

arXiv2026

Entropy-Adaptive Fine-Tuning: Resolving Confident Conflicts to Mitigate Forgetting

Supervised Fine-Tuning (SFT) is the standard paradigm for domain adaptation, yet it frequently incurs the cost of catastrophic forgetting. In sharp contrast, on-policy Reinforcement Learning (RL) effectively preserves general capabilities. We investigate this discrepancy and identify a fundamental distributional gap: while RL aligns with the model's internal belief, SFT forces the model to fit external supervision. This mismatch often manifests as "Confident Conflicts" tokens characterized by low probability but low entropy. In these instances, the model is highly confident in its own prediction but is forced to learn a divergent ground truth, triggering destructive gradient updates. To address this, we propose Entropy-Adaptive Fine-Tuning (EAFT). Unlike methods relying solely on prediction probability, EAFT utilizes token-level entropy as a gating mechanism to distinguish between epistemic uncertainty and knowledge conflict. This allows the model to learn from uncertain samples while suppressing gradients on conflicting data. Extensive experiments on Qwen and GLM series (ranging from 4B to 32B parameters) across mathematical, medical, and agentic domains confirm our hypothesis. EAFT consistently matches the downstream performance of standard SFT while significantly mitigating the degradation of general capabilities.

cs.LG↗

arXiv2026

BARE: Towards Bias-Aware and Reasoning-Enhanced One-Tower Visual Grounding

Visual Grounding (VG), which aims to locate a specific region referred to by expressions, is a fundamental yet challenging task in the multimodal understanding fields. While recent grounding transfer works have advanced the field through one-tower architectures, they still suffer from two primary limitations: (1) over-entangled multimodal representations that exacerbate deceptive modality biases, and (2) insufficient semantic reasoning that hinders the comprehension of referential cues. In this paper, we propose BARE, a bias-aware and reasoning-enhanced framework for one-tower visual grounding. BARE introduces a mechanism that preserves modality-specific features and constructs referential semantics through three novel modules: (i) language salience modulator, (ii) visual bias correction and (iii) referential relationship enhancement, which jointly mitigate multimodal distractions and enhance referential comprehension. Extensive experimental results on five benchmarks demonstrate that BARE not only achieves state-of-the-art performance but also delivers superior computational efficiency compared to existing approaches. The code is publicly accessible at https://github.com/Marloweeee/BARE.

cs.CV↗

arXiv2025

FourierSR: A Fourier Token-based Plugin for Efficient Image Super-Resolution

Image super-resolution (SR) aims to recover low-resolution images to high-resolution images, where improving SR efficiency is a high-profile challenge. However, commonly used units in SR, like convolutions and window-based Transformers, have limited receptive fields, making it challenging to apply them to improve SR under extremely limited computational cost. To address this issue, inspired by modeling convolution theorem through token mix, we propose a Fourier token-based plugin called FourierSR to improve SR uniformly, which avoids the instability or inefficiency of existing token mix technologies when applied as plug-ins. Furthermore, compared to convolutions and windows-based Transformers, our FourierSR only utilizes Fourier transform and multiplication operations, greatly reducing complexity while having global receptive fields. Experimental results show that our FourierSR as a plug-and-play unit brings an average PSNR gain of 0.34dB for existing efficient SR methods on Manga109 test set at the scale of x4, while the average increase in the number of Params and FLOPs is only 0.6% and 1.5% of original sizes. We will release our codes upon acceptance.

cs.CV↗

arXiv2025

Seeing Through the Rain: Resolving High-Frequency Conflicts in Deraining and Super-Resolution via Diffusion Guidance

Clean images are crucial for visual tasks such as small object detection, especially at high resolutions. However, real-world images are often degraded by adverse weather, and weather restoration methods may sacrifice high-frequency details critical for analyzing small objects. A natural solution is to apply super-resolution (SR) after weather removal to recover both clarity and fine structures. However, simply cascading restoration and SR struggle to bridge their inherent conflict: removal aims to remove high-frequency weather-induced noise, while SR aims to hallucinate high-frequency textures from existing details, leading to inconsistent restoration contents. In this paper, we take deraining as a case study and propose DHGM, a Diffusion-based High-frequency Guided Model for generating clean and high-resolution images. DHGM integrates pre-trained diffusion priors with high-pass filters to simultaneously remove rain artifacts and enhance structural details. Extensive experiments demonstrate that DHGM achieves superior performance over existing methods, with lower costs.

cs.CV↗

arXiv2025

SpecGen: Neural Spectral BRDF Generation via Spectral-Spatial Tri-plane Aggregation

Synthesizing spectral images across different wavelengths is essential for photorealistic rendering. Unlike conventional spectral uplifting methods that convert RGB images into spectral ones, we introduce SpecGen, a novel method that generates spectral bidirectional reflectance distribution functions (BRDFs) from a single RGB image of a sphere. This enables spectral image rendering under arbitrary illuminations and shapes covered by the corresponding material. A key challenge in spectral BRDF generation is the scarcity of measured spectral BRDF data. To address this, we propose the Spectral-Spatial Tri-plane Aggregation (SSTA) network, which models reflectance responses across wavelengths and incident-outgoing directions, allowing the training strategy to leverage abundant RGB BRDF data to enhance spectral BRDF generation. Experiments show that our method accurately reconstructs spectral BRDFs from limited spectral data and surpasses state-of-the-art methods in hyperspectral image reconstruction, achieving an improvement of 8 dB in PSNR. Codes and data will be released upon acceptance.

cs.CV↗

arXiv2025

Engineering the localization transition in a Charge-Kondo circuit

Charge Kondo circuits consist of metallic islands connected by single-mode quantum point contacts (QPCs). The island's charging energy makes these circuits tunable quantum simulators of various strongly interacting models. Here we propose a circuit that realizes the Kondo effect with effective Luttinger-liquid interactions, and show that it undergoes a localization transition in which the QPC transmission is fully suppressed below a critical value. Experimental signatures include a diverging charge susceptibility and an entropy step. Our findings open a path toward realizing localization transitions in more exotic settings.

cond-mat.mes-hall↗

arXiv2025

Measurement-Constrained Sampling for Text-Prompted Blind Face Restoration

Blind face restoration (BFR) may correspond to multiple plausible high-quality (HQ) reconstructions under extremely low-quality (LQ) inputs. However, existing methods typically produce deterministic results, struggling to capture this one-to-many nature. In this paper, we propose a Measurement-Constrained Sampling (MCS) approach that enables diverse LQ face reconstructions conditioned on different textual prompts. Specifically, we formulate BFR as a measurement-constrained generative task by constructing an inverse problem through controlled degradations of coarse restorations, which allows posterior-guided sampling within text-to-image diffusion. Measurement constraints include both Forward Measurement, which ensures results align with input structures, and Reverse Measurement, which produces projection spaces, ensuring that the solution can align with various prompts. Experiments show that our MCS can generate prompt-aligned results and outperforms existing BFR methods. Codes will be released after acceptance.

cs.CV↗

arXiv2025

VoiceAgentEval: A Dual-Dimensional Benchmark for Expert-Level Intelligent Voice-Agent Evaluation of Xbench's Professional-Aligned Series

We propose OutboundEval, a comprehensive benchmark for evaluating large language models (LLMs) in expert-level intelligent outbound calling scenarios. Unlike existing methods that suffer from three key limitations - insufficient dataset diversity and category coverage, unrealistic user simulation, and inaccurate evaluation metrics - OutboundEval addresses these issues through a structured framework. First, we design a benchmark spanning six major business domains and 30 representative sub-scenarios, each with scenario-specific process decomposition, weighted scoring, and domain-adaptive metrics. Second, we develop a large-model-driven User Simulator that generates diverse, persona-rich virtual users with realistic behaviors, emotional variability, and communication styles, providing a controlled yet authentic testing environment. Third, we introduce a dynamic evaluation method that adapts to task variations, integrating automated and human-in-the-loop assessment to measure task execution accuracy, professional knowledge application, adaptability, and user experience quality. Experiments on 12 state-of-the-art LLMs reveal distinct trade-offs between expert-level task completion and interaction fluency, offering practical insights for building reliable, human-like outbound AI systems. OutboundEval establishes a practical, extensible, and domain-oriented standard for benchmarking LLMs in professional applications.

cs.AI↗

arXiv2025

Self-Supervised Selective-Guided Diffusion Model for Old-Photo Face Restoration

Old-photo face restoration poses significant challenges due to compounded degradations such as breakage, fading, and severe blur. Existing pre-trained diffusion-guided methods either rely on explicit degradation priors or global statistical guidance, which struggle with localized artifacts or face color. We propose Self-Supervised Selective-Guided Diffusion (SSDiff), which leverages pseudo-reference faces generated by a pre-trained diffusion model under weak guidance. These pseudo-labels exhibit structurally aligned contours and natural colors, enabling region-specific restoration via staged supervision: structural guidance applied throughout the denoising process and color refinement in later steps, aligned with the coarse-to-fine nature of diffusion. By incorporating face parsing maps and scratch masks, our method selectively restores breakage regions while avoiding identity mismatch. We further construct VintageFace, a 300-image benchmark of real old face photos with varying degradation levels. SSDiff outperforms existing GAN-based and diffusion-based methods in perceptual quality, fidelity, and regional controllability. Code link: https://github.com/PRIS-CV/SSDiff.

cs.CV↗

arXiv2025

AutoRed: A Free-form Adversarial Prompt Generation Framework for Automated Red Teaming

The safety of Large Language Models (LLMs) is crucial for the development of trustworthy AI applications. Existing red teaming methods often rely on seed instructions, which limits the semantic diversity of the synthesized adversarial prompts. We propose AutoRed, a free-form adversarial prompt generation framework that removes the need for seed instructions. AutoRed operates in two stages: (1) persona-guided adversarial instruction generation, and (2) a reflection loop to iteratively refine low-quality prompts. To improve efficiency, we introduce a verifier to assess prompt harmfulness without querying the target models. Using AutoRed, we build two red teaming datasets -- AutoRed-Medium and AutoRed-Hard -- and evaluate eight state-of-the-art LLMs. AutoRed achieves higher attack success rates and better generalization than existing baselines. Our results highlight the limitations of seed-based approaches and demonstrate the potential of free-form red teaming for LLM safety evaluation. We will open source our datasets in the near future.

cs.CL↗

arXiv2025

Quantum work statistics across a critical point: full crossover from sudden quench to the adiabatic limit

When an external parameter drives a system across a quantum phase transition at a finite rate, work is performed on the system and entropy is dissipated, due to the creation of excitations via the Kibble-Zurek mechanism. Although both the adiabatic and sudden-quench limits have been studied in detail, the quantum work statistics along the crossover connecting these limits has largely been an open question. Here we obtain exact scaling functions for the work statistics along the full crossover from adiabatic to sudden-quench limits for critical quantum impurity problems, by combining linear response theory, conformal field theory, and the numerical renormalization group. These predictions can be tested in charge-multichannel Kondo quantum dot devices, where the dissipated work corresponds to the creation of nontrivial excitations such as Majorana fermions or Fibonacci anyons.

quant-ph↗

arXiv2025

PolGS: Polarimetric Gaussian Splatting for Fast Reflective Surface Reconstruction

Efficient shape reconstruction for surfaces with complex reflectance properties is crucial for real-time virtual reality. While 3D Gaussian Splatting (3DGS)-based methods offer fast novel view rendering by leveraging their explicit surface representation, their reconstruction quality lags behind that of implicit neural representations, particularly in the case of recovering surfaces with complex reflective reflectance. To address these problems, we propose PolGS, a Polarimetric Gaussian Splatting model allowing fast reflective surface reconstruction in 10 minutes. By integrating polarimetric constraints into the 3DGS framework, PolGS effectively separates specular and diffuse components, enhancing reconstruction quality for challenging reflective materials. Experimental results on the synthetic and real-world dataset validate the effectiveness of our method.

cs.CV↗

arXiv2025

Controllable-Continuous Color Editing in Diffusion Model via Color Mapping

In recent years, text-driven image editing has made significant progress. However, due to the inherent ambiguity and discreteness of natural language, color editing still faces challenges such as insufficient precision and difficulty in achieving continuous control. Although linearly interpolating the embedding vectors of different textual descriptions can guide the model to generate a sequence of images with varying colors, this approach lacks precise control over the range of color changes in the output images. Moreover, the relationship between the interpolation coefficient and the resulting image color is unknown and uncontrollable. To address these issues, we introduce a color mapping module that explicitly models the correspondence between the text embedding space and image RGB values. This module predicts the corresponding embedding vector based on a given RGB value, enabling precise color control of the generated images while maintaining semantic consistency. Users can specify a target RGB range to generate images with continuous color variations within the desired range, thereby achieving finer-grained, continuous, and controllable color editing. Experimental results demonstrate that our method performs well in terms of color continuity and controllability.

cs.CV↗

arXiv2025

Realizing the interacting resonant level model using a quantum dot detector

The interacting resonant level model (IRLM) is the simplest quantum impurity model to display strongly correlated effects in mesoscopic systems, which triggered its extensive theoretical study. However, to date, there have not been any realizations of the model with controllable interaction parameter, and thus the detailed predictions could not be confirmed. Here we use a recently developed approach to Anderson orthogonality catastrophe physics, using a charge detector coupled to a quantum dot (QD) system, to devise a simple experimental system which could display IRLM behavior and detail its predictions. At the same time, the mapping to IRLM allows us to determine the interaction parameter of the charge detector using simple experimental probes.

cond-mat.mes-hall↗

arXiv2025

Dual-domain Modulation Network for Lightweight Image Super-Resolution

Lightweight image super-resolution (SR) aims to reconstruct high-resolution images from low-resolution images under limited computational costs. We find that existing frequency-based SR methods cannot balance the reconstruction of overall structures and high-frequency parts. Meanwhile, these methods are inefficient for handling frequency features and unsuitable for lightweight SR. In this paper, we show that introducing both wavelet and Fourier information allows our model to consider both high-frequency features and overall SR structure reconstruction while reducing costs. Specifically, we propose a Dual-domain Modulation Network that integrates both wavelet and Fourier information for enhanced frequency modeling. Unlike existing methods that rely on a single frequency representation, our design combines wavelet-domain modulation via a Wavelet-domain Modulation Transformer (WMT) with global Fourier supervision, enabling complementary spectral learning well-suited for lightweight SR. Experimental results show that our method achieves a comparable PSNR to SRFormer and MambaIR while with less than 50\% and 60\% of their FLOPs and achieving inference speeds 15.4x and 5.4x faster, respectively, demonstrating the effectiveness of our method on SR quality and lightweight. Code link: https://github.com/24wenjie-li/DMNet

eess.IV↗