arXiv subjects

Jiale Yuan

Publications and source records attributed to Jiale Yuan.

15 recordsLinked to original sources

arXiv2025

Steering One-Step Diffusion Model with Fidelity-Rich Decoder for Fast Image Compression

Diffusion-based image compression has demonstrated impressive perceptual performance. However, it suffers from two critical drawbacks: (1) excessive decoding latency due to multi-step sampling, and (2) poor fidelity resulting from over-reliance on generative priors. To address these issues, we propose SODEC, a novel single-step diffusion image compression model. We argue that in image compression, a sufficiently informative latent renders multi-step refinement unnecessary. Based on this insight, we leverage a pre-trained VAE-based model to produce latents with rich information, and replace the iterative denoising process with a single-step decoding. Meanwhile, to improve fidelity, we introduce the fidelity guidance module, encouraging output that is faithful to the original image. Furthermore, we design the rate annealing training strategy to enable effective training under extremely low bitrates. Extensive experiments show that SODEC significantly outperforms existing methods, achieving superior rate-distortion-perception performance. Moreover, compared to previous diffusion-based compression models, SODEC improves decoding speed by more than 20$\times$. Code is released at: https://github.com/zhengchen1999/SODEC.

cs.CV↗

arXiv2025

Training-free Generation of Temporally Consistent Rewards from VLMs

Recent advances in vision-language models (VLMs) have significantly improved performance in embodied tasks such as goal decomposition and visual comprehension. However, providing accurate rewards for robotic manipulation without fine-tuning VLMs remains challenging due to the absence of domain-specific robotic knowledge in pre-trained datasets and high computational costs that hinder real-time applicability. To address this, we propose $\mathrm{T}^2$-VLM, a novel training-free, temporally consistent framework that generates accurate rewards through tracking the status changes in VLM-derived subgoals. Specifically, our method first queries the VLM to establish spatially aware subgoals and an initial completion estimate before each round of interaction. We then employ a Bayesian tracking algorithm to update the goal completion status dynamically, using subgoal hidden states to generate structured rewards for reinforcement learning (RL) agents. This approach enhances long-horizon decision-making and improves failure recovery capabilities with RL. Extensive experiments indicate that $\mathrm{T}^2$-VLM achieves state-of-the-art performance in two robot manipulation benchmarks, demonstrating superior reward accuracy with reduced computation consumption. We believe our approach not only advances reward generation techniques but also contributes to the broader field of embodied AI. Project website: https://t2-vlm.github.io/.

cs.RO↗

arXiv2025

Self-NPO: Data-Free Diffusion Model Enhancement via Truncated Diffusion Fine-Tuning

Diffusion models have demonstrated remarkable success in various visual generation tasks, including image, video, and 3D content generation. Preference optimization (PO) is a prominent and growing area of research that aims to align these models with human preferences. While existing PO methods primarily concentrate on producing favorable outputs, they often overlook the significance of classifier-free guidance (CFG) in mitigating undesirable results. Diffusion-NPO addresses this gap by introducing negative preference optimization (NPO), training models to generate outputs opposite to human preferences and thereby steering them away from unfavorable outcomes through CFG. However, prior NPO approaches rely on costly and fragile procedures for obtaining explicit preference annotations (e.g., manual pairwise labeling or reward model training), limiting their practicality in domains where such data are scarce or difficult to acquire. In this work, we propose Self-NPO, specifically truncated diffusion fine-tuning, a data-free approach of negative preference optimization by directly learning from the model itself, eliminating the need for manual data labeling or reward model training. This data-free approach is highly efficient (less than 1% training cost of Diffusion-NPO) and achieves comparable performance to Diffusion-NPO in a data-free manner. We demonstrate that Self-NPO integrates seamlessly into widely used diffusion models, including SD1.5, SDXL, and CogVideoX, as well as models already optimized for human preferences, consistently enhancing both their generation quality and alignment with human preferences. Code is available at https://github.com/G-U-N/Diffusion-NPO.

cs.CV↗

arXiv2025

"See the World, Discover Knowledge": A Chinese Factuality Evaluation for Large Vision Language Models

The evaluation of factual accuracy in large vision language models (LVLMs) has lagged behind their rapid development, making it challenging to fully reflect these models' knowledge capacity and reliability. In this paper, we introduce the first factuality-based visual question-answering benchmark in Chinese, named ChineseSimpleVQA, aimed at assessing the visual factuality of LVLMs across 8 major topics and 56 subtopics. The key features of this benchmark include a focus on the Chinese language, diverse knowledge types, a multi-hop question construction, high-quality data, static consistency, and easy-to-evaluate through short answers. Moreover, we contribute a rigorous data construction pipeline and decouple the visual factuality into two parts: seeing the world (i.e., object recognition) and discovering knowledge. This decoupling allows us to analyze the capability boundaries and execution mechanisms of LVLMs. Subsequently, we evaluate 34 advanced open-source and closed-source models, revealing critical performance gaps within this field. Our evaluation-friendly code and data have already been open-sourced.

cs.CL↗

arXiv2025

Global Compression Commander: Plug-and-Play Inference Acceleration for High-Resolution Large Vision-Language Models

Large vision-language models (LVLMs) excel at visual understanding, but face efficiency challenges due to quadratic complexity in processing long multi-modal contexts. While token compression can reduce computational costs, existing approaches are designed for single-view LVLMs and fail to consider the unique multi-view characteristics of high-resolution LVLMs with dynamic cropping. Existing methods treat all tokens uniformly, but our analysis reveals that global thumbnails can naturally guide the compression of local crops by providing holistic context for informativeness evaluation. In this paper, we first analyze dynamic cropping strategy, revealing both the complementary nature between thumbnails and crops, and the distinctive characteristics across different crops. Based on our observations, we propose ``Global Compression Commander'' (\textit{i.e.}, \textbf{GlobalCom$^2$}), a novel plug-and-play token compression framework for HR-LVLMs. GlobalCom$^2$ leverages thumbnail as the ``commander'' to guide the compression of local crops, adaptively preserving informative details while eliminating redundancy. Extensive experiments show that GlobalCom$^2$ maintains over \textbf{90\%} performance while compressing \textbf{90\%} visual tokens, reducing FLOPs and peak memory to \textbf{9.1\%} and \textbf{60\%}.

cs.CV↗

arXiv2024

LongDocURL: a Comprehensive Multimodal Long Document Benchmark Integrating Understanding, Reasoning, and Locating

Large vision language models (LVLMs) have improved the document understanding capabilities remarkably, enabling the handling of complex document elements, longer contexts, and a wider range of tasks. However, existing document understanding benchmarks have been limited to handling only a small number of pages and fail to provide a comprehensive analysis of layout elements locating. In this paper, we first define three primary task categories: Long Document Understanding, numerical Reasoning, and cross-element Locating, and then propose a comprehensive benchmark, LongDocURL, integrating above three primary tasks and comprising 20 sub-tasks categorized based on different primary tasks and answer evidences. Furthermore, we develop a semi-automated construction pipeline and collect 2,325 high-quality question-answering pairs, covering more than 33,000 pages of documents, significantly outperforming existing benchmarks. Subsequently, we conduct comprehensive evaluation experiments on both open-source and closed-source models across 26 different configurations, revealing critical performance gaps in this field.

cs.AI↗

arXiv2024

Ideal flat and resolved SU(3) Landau levels in three dimensions

Landau levels (LLs) are of great importance for understanding the quantum Hall effect and associated many-body physics. Recently, their three-dimensional (3D) counterparts, i.e., dispersionless 3D LLs with well-defined quantum numbers, have attracted significant attention but have not yet been reported. Here we theoretically propose and experimentally observe 3D LLs with a sharply quantized spectrum in a diamond acoustic lattice, where the eigenstates are characterized by SU(3) quantum numbers. The engineered inhomogeneous hopping strengths not only introduce pseudomagnetic fields that quantize the nodal lines into LLs but also provide three bosonic degrees of freedom, embedding a generic SU(3) symmetry into the LLs. Using a phased array of acoustic sources, we selectively excite distinct eigenstates within the degenerate LL multiplets and visualize their 3D eigenmodes. Importantly, our approach enables the precise reconstruction of SU(3) quantum numbers directly from eigenmode correlations. Our results establish SU(3) LLs as a tractable model in artificial platforms, and pave the way for synthesizing LLs with zero dispersion and countable quantum numbers in arbitrary dimensions.

cond-mat.mes-hall↗

arXiv2024

ConvLLaVA: Hierarchical Backbones as Visual Encoder for Large Multimodal Models

High-resolution Large Multimodal Models (LMMs) encounter the challenges of excessive visual tokens and quadratic visual complexity. Current high-resolution LMMs address the quadratic complexity while still generating excessive visual tokens. However, the redundancy in visual tokens is the key problem as it leads to more substantial compute. To mitigate this issue, we propose ConvLLaVA, which employs ConvNeXt, a hierarchical backbone, as the visual encoder of LMM to replace Vision Transformer (ViT). ConvLLaVA compresses high-resolution images into information-rich visual features, effectively preventing the generation of excessive visual tokens. To enhance the capabilities of ConvLLaVA, we propose two critical optimizations. Since the low-resolution pretrained ConvNeXt underperforms when directly applied on high resolution, we update it to bridge the gap. Moreover, since ConvNeXt's original compression ratio is inadequate for much higher resolution inputs, we train a successive stage to further compress the visual tokens, thereby reducing redundancy. These optimizations enable ConvLLaVA to support inputs of 1536x1536 resolution generating only 576 visual tokens, capable of handling images of arbitrary aspect ratios. Experimental results demonstrate that our method achieves competitive performance with state-of-the-art models on mainstream benchmarks. The ConvLLaVA model series are publicly available at https://github.com/alibaba/conv-llava.

cs.CV↗

arXiv2024

Testing the unified bounds of quantum speed limit

Quantum speed limits (QSLs) impose fundamental constraints on the evolution speed of quantum systems. Traditionally, the Mandelstam-Tamm (MT) and Margolus-Levitin (ML) bounds have been widely employed, relying on the standard deviation and mean of energy distribution to define the QSLs. However, these universal bounds only offer loose restrictions on the quantum evolution. Here we introduce the generalized ML bounds, which prove to be more stringent in constraining dynamic evolution, by utilizing moments of energy spectra of arbitrary orders, even noninteger orders. To validate our findings, we conduct experiments in a superconducting circuit, where we have the capability to prepare a wide range of quantum photonic states and rigorously test these bounds by measuring the evolution of the system and its photon statistics using quantum state tomography. While, in general, the MT bound is effective for short-time evolution, we identify specific parameter regimes where either the MT or the generalized ML bounds suffice to constrain the entire evolution. Our findings not only establish new criteria for estimating QSLs but also substantially enhance our comprehension of the dynamic evolution of quantum systems.

quant-ph↗

arXiv2024

Realization of all-band-flat photonic lattices

Flatbands play an important role in correlated quantum matter and have novel applications in photonic lattices. Synthetic magnetic fields and destructive interference in lattices are traditionally used to obtain flatbands. However, such methods can only obtain a few flatbands with most bands remaining dispersive. Here we realize all-band-flat photonic lattices of an arbitrary size by precisely controlling the coupling strengths between lattice sites to mimic those in Fock-state lattices. This allows us to go beyond the perturbative regime of strain engineering and group all eigenmodes in flatbands, which simultaneously achieves high band flatness and large usable bandwidth. We map out the distribution of each flatband in the lattices and selectively excite the eigenmodes with different chiralities. Our method paves a new way in controlling band structure and topology of photonic lattices.

physics.optics↗

arXiv2022

Coherent control of quantum topological states of light in Fock-state lattices

Topological photonics provides a novel platform to explore topological physics beyond traditional electronic materials and stimulates promising applications in topologically protected light transport and lasers. Classical degrees of freedom such as polarizations and wavevectors are routinely used to synthesize topological light modes. Beyond the classical regime, inherent quantum nature of light gives birth to a wealth of fundamentally distinct topological states, which offer topological protection in quantum information processing. Here we implement such experiments on topological states of quantized light in a superconducting circuit, on which three resonators are tunably coupled to a gmon qubit. We construct one and two-dimensional Fock-state lattices where topological transport of zero-energy states, strain induced pseudo-Landau levels, valley Hall effect and Haldane chiral edge currents are demonstrated. Our study extends the topological states of light to the quantum regime, bridges topological phases of condensed matter physics with circuit quantum electrodynamics, and offers a new freedom in controlling the quantum states of multiple resonators.

quant-ph↗

arXiv2021

Synthesizing five-body interaction in a superconducting quantum circuit

Synthesizing many-body interaction Hamiltonian is a central task in quantum simulation. However, it is challenging to synthesize interactions including more than two spins. Borrowing tools from quantum optics, we synthesize five-body spin-exchange interaction in a superconducting quantum circuit by simultaneously exciting four independent qubits with time-energy correlated photon quadruples generated from a qudit. During the dynamic evolution of the five-body interaction, a Greenberger-Horne-Zeilinger state is generated in a single step with fidelity estimated to be $0.685$. We compare the influence of noise on the three-, four- and five-body interaction as a step toward answering the question on the quantum origin of chiral molecules. We also demonstrate a many-body Mach-Zehnder interferometer which potentially has a Heisenberg-limit sensitivity. This study paves a way for quantum simulation involving many-body interactions and high excited states of quantum circuits.

quant-ph↗

arXiv2021

Unification of valley and anomalous Hall effects in a strained lattice

Two dimensional lattices are an important stage for studying many aspects of quantum physics, in particular the topological phases. The valley Hall and anomalous Hall effects are two representative topological phenomena. Here we show that they can be unified in a strained honeycomb lattice, where the hopping strengths between neighboring sites are designed by mimicking those between the Fock states in a three-mode Jaynes-Cummings model. Such a strain induces an effective magnetic field which results in quantized Landau levels. The eigenstates in the zeroth Landau level can be represented by the eigenstates of a large pseudo-spin. We find that the valley Hall current and the chiral edge current in the Haldane model correspond to the spin precession around different axes. Our study sheds light on connection between seemingly unrelated topological phases in condensed matter physics.

cond-mat.mes-hall↗

arXiv2020

Electric-field-induced modulation of thermal conductivity in poly(vinylidene fluoride)

Phonon engineering focuses on heat transport modulation on atomic-scale. Different from reported methods, it is shown that electric field can also modulate heat transport in ferroelectric polymers, poly(vinylidene fluoride), by both simulation and measurement. Interestingly, thermal conductivities of poly(vinylidene fluoride) array can be enhanced by a factor of 3.25 along the polarization direction by simulation. The semi-crystalline poly(vinylidene fluoride) film can be also enhanced by a factor of 1.5 which is found by both simulation and measurement. The morphology and phonon property analysis reveal that the enhancement arises from the higher inter-chain lattice order, stronger inter-chain interaction, higher phonon group velocity and suppressed phonon scattering. This study offers a new modulation strategy with quick response and without fillers.

cond-mat.mes-hall↗

arXiv2019

Thermal Boundary Resistance Measurement and Analysis Across SiC/SiO2 Interface

Silicon Carbide (SiC) is a typical material for third-generation semiconductor. The thermal boundary resistance (TBR) of 4H-SiC/SiO2 interface, was investigated by both experimental measurements and theoretical calculations. The structure of 4H-SiC/SiO2 was characterized by using transmission electron microscopy and X-ray diffraction. The TBR is measured as 8.11*10-8 m2K/W by 3-omega method. Furthermore, the diffuse mismatch model was employed to predict the TBR of different interfaces which is in good agreement with measurements. Heat transport behavior based on phonon scattering perspective was also discussed to understand the variations of TBR across different interfaces. Besides, the intrinsic thermal conductivity of SiO2 thin films (200~1,500 nm in thickness) on 4H-SiC substrates was measured by 3 omega procedure, as 1.42 W/mK at room temperature. It is believed the presented results could provide useful insights on the thermal management and heat dissipation for SiC devices.

physics.app-ph↗