arXiv subjects

Liming Zhao

Publications and source records attributed to Liming Zhao.

At least 19 recordsLinked to original sources

arXiv2025

Hybrid-Level Instruction Injection for Video Token Compression in Multi-modal Large Language Models

Recent Multi-modal Large Language Models (MLLMs) have been challenged by the computational overhead resulting from massive video frames, often alleviated through compression strategies. However, the visual content is not equally contributed to user instructions, existing strategies (\eg, average pool) inevitably lead to the loss of potentially useful information. To tackle this, we propose the Hybrid-level Instruction Injection Strategy for Conditional Token Compression in MLLMs (HICom), utilizing the instruction as a condition to guide the compression from both local and global levels. This encourages the compression to retain the maximum amount of user-focused information while reducing visual tokens to minimize computational burden. Specifically, the instruction condition is injected into the grouped visual tokens at the local level and the learnable tokens at the global level, and we conduct the attention mechanism to complete the conditional compression. From the hybrid-level compression, the instruction-relevant visual parts are highlighted while the temporal-spatial structure is also preserved for easier understanding of LLMs. To further unleash the potential of HICom, we introduce a new conditional pre-training stage with our proposed dataset HICom-248K. Experiments show that our HICom can obtain distinguished video understanding ability with fewer tokens, increasing the performance by 2.43\% average on three multiple-choice QA benchmarks and saving 78.8\% tokens compared with the SOTA method. The code is available at https://github.com/lntzm/HICom.

cs.CV↗

arXiv2025

Rethinking Video Tokenization: A Conditioned Diffusion-based Approach

Existing video tokenizers typically use the traditional Variational Autoencoder (VAE) architecture for video compression and reconstruction. However, to achieve good performance, its training process often relies on complex multi-stage training tricks that go beyond basic reconstruction loss and KL regularization. Among these tricks, the most challenging is the precise tuning of adversarial training with additional Generative Adversarial Networks (GANs) in the final stage, which can hinder stable convergence. In contrast to GANs, diffusion models offer more stable training processes and can generate higher-quality results. Inspired by these advantages, we propose CDT, a novel Conditioned Diffusion-based video Tokenizer, that replaces the GAN-based decoder with a conditional causal diffusion model. The encoder compresses spatio-temporal information into compact latents, while the decoder reconstructs videos through a reverse diffusion process conditioned on these latents. During inference, we incorporate a feature cache mechanism to generate videos of arbitrary length while maintaining temporal continuity and adopt sampling acceleration technique to enhance efficiency. Trained using only a basic MSE diffusion loss for reconstruction, along with KL term and LPIPS perceptual loss from scratch, extensive experiments demonstrate that CDT achieves state-of-the-art performance in video reconstruction tasks with just a single-step sampling. Even a scaled-down version of CDT (3$\times$ inference speedup) still performs comparably with top baselines. Moreover, the latent video generation model trained with CDT also exhibits superior performance. The source code and pretrained weights are available at https://github.com/ali-vilab/CDT.

cs.CV↗

arXiv2025

Dimension Effect of Nanocarbon Precursors on Diamond Synthesis and Transformation Mechanism under Extreme Conditions

Diamond holds significant promise for a wide range of applications due to its exceptional physicochemical properties. Investigating the controlled diamond preparation from nanocarbon precursors with varying dimensions is crucial to optimize the transition conditions and even elucidate the daunting transformation mechanism, however, this remains outstanding challenge despite considerable effort. Herein, we report the imperative dimension effect of nanocarbon precursors on diamond synthesis and physical mechanism under high temperature and high pressure, by comparing the distinct transition processes of zero-dimensional (0D) carbon nanocages (CNCs) and one-dimensional (1D) carbon nanotubes (CNTs) from conventional graphite. The optical and structural characterizations evidently demonstrated that both 0D CNCs and 1D CNTs first undergo collapse and graphitization, followed by the formation of mixed amorphous carbon with embedded diamond clusters, eventually leading to cubic diamond. The plotted pressure-temperature diagram exhibits the unique dimension effect of carbon nanomaterials to diamond transformation. These results provide valuable insights into the phase transition mechanisms of diamond synthesis and its derivatives under extreme conditions.

cond-mat.mtrl-sci↗

arXiv2024

ContextHOI: Spatial Context Learning for Human-Object Interaction Detection

Spatial contexts, such as the backgrounds and surroundings, are considered critical in Human-Object Interaction (HOI) recognition, especially when the instance-centric foreground is blurred or occluded. Recent advancements in HOI detectors are usually built upon detection transformer pipelines. While such an object-detection-oriented paradigm shows promise in localizing objects, its exploration of spatial context is often insufficient for accurately recognizing human actions. To enhance the capabilities of object detectors for HOI detection, we present a dual-branch framework named ContextHOI, which efficiently captures both object detection features and spatial contexts. In the context branch, we train the model to extract informative spatial context without requiring additional hand-craft background labels. Furthermore, we introduce context-aware spatial and semantic supervision to the context branch to filter out irrelevant noise and capture informative contexts. ContextHOI achieves state-of-the-art performance on the HICO-DET and v-coco benchmarks. For further validation, we construct a novel benchmark, HICO-ambiguous, which is a subset of HICO-DET that contains images with occluded or impaired instance cues. Extensive experiments across all benchmarks, complemented by visualizations, underscore the enhancements provided by ContextHOI, especially in recognizing interactions involving occluded or blurred instances.

cs.CV↗

arXiv2024

Orchestrating the Symphony of Prompt Distribution Learning for Human-Object Interaction Detection

Human-object interaction (HOI) detectors with popular query-transformer architecture have achieved promising performance. However, accurately identifying uncommon visual patterns and distinguishing between ambiguous HOIs continue to be difficult for them. We observe that these difficulties may arise from the limited capacity of traditional detector queries in representing diverse intra-category patterns and inter-category dependencies. To address this, we introduce the Interaction Prompt Distribution Learning (InterProDa) approach. InterProDa learns multiple sets of soft prompts and estimates category distributions from various prompts. It then incorporates HOI queries with category distributions, making them capable of representing near-infinite intra-category dynamics and universal cross-category relationships. Our InterProDa detector demonstrates competitive performance on HICO-DET and vcoco benchmarks. Additionally, our method can be integrated into most transformer-based HOI detectors, significantly enhancing their performance with minimal additional parameters.

cs.CV↗

arXiv2024

Improved Video VAE for Latent Video Diffusion Model

Variational Autoencoder (VAE) aims to compress pixel data into low-dimensional latent space, playing an important role in OpenAI's Sora and other latent video diffusion generation models. While most of existing video VAEs inflate a pretrained image VAE into the 3D causal structure for temporal-spatial compression, this paper presents two astonishing findings: (1) The initialization from a well-trained image VAE with the same latent dimensions suppresses the improvement of subsequent temporal compression capabilities. (2) The adoption of causal reasoning leads to unequal information interactions and unbalanced performance between frames. To alleviate these problems, we propose a keyframe-based temporal compression (KTC) architecture and a group causal convolution (GCConv) module to further improve video VAE (IV-VAE). Specifically, the KTC architecture divides the latent space into two branches, in which one half completely inherits the compression prior of keyframes from a lower-dimension image VAE while the other half involves temporal compression through 3D group causal convolution, reducing temporal-spatial conflicts and accelerating the convergence speed of video VAE. The GCConv in above 3D half uses standard convolution within each frame group to ensure inter-frame equivalence, and employs causal logical padding between groups to maintain flexibility in processing variable frame video. Extensive experiments on five benchmarks demonstrate the SOTA video reconstruction and generation capabilities of the proposed IV-VAE (https://wpy1999.github.io/IV-VAE/).

cs.CV↗

arXiv2024

Control System Design and Experiments for Autonomous Underwater Helicopter Docking Procedure Based on Acoustic-inertial-optical Guidance

A control system structure for the underwater docking procedure of an Autonomous Underwater Helicopter (AUH) is proposed in this paper, which utilizes acoustic-inertial-optical guidance. Unlike conventional Autonomous Underwater Vehicles (AUVs), the maneuverability requirements for AUHs are more stringent during the docking procedure, requiring it to remain stationary or have minimal horizontal movement while moving vertically. The docking procedure is divided into two stages: Homing and Landing, each stage utilizing different guidance methods. Additionally, a segmented aligning strategy operating at various altitudes and a linear velocity decision are both adopted in Landing stage. Due to the unique structure of the Subsea Docking System (SDS), the AUH is required to dock onto the SDS in a fixed orientation with specific attitude and altitude. Therefore, a particular criterion is proposed to determine whether the AUH has successfully docked onto the SDS. Furthermore, the effectiveness and robustness of the proposed control method in AUH's docking procedure are demonstrated through pool experiments and sea trials.

cs.RO↗

arXiv2023

Learning Restricted Boltzmann Machines with greedy quantum search

Restricted Boltzmann Machines (RBMs) are widely used probabilistic undirected graphical models with visible and latent nodes, playing an important role in statistics and machine learning. The task of structure learning for RBMs involves inferring the underlying graph by using samples from the visible nodes. Specifically, learning the two-hop neighbors of each visible node allows for the inference of the graph structure. Prior research has addressed the structure learning problem for specific classes of RBMs, namely ferromagnetic and locally consistent RBMs. In this paper, we extend the scope to the quantum computing domain and propose corresponding quantum algorithms for this problem. Our study demonstrates that the proposed quantum algorithms yield a polynomial speedup compared to the classical algorithms for learning the structure of these two classes of RBMs.

quant-ph↗

arXiv2023

Provable learning of quantum states with graphical models

The complete learning of an $n$-qubit quantum state requires samples exponentially in $n$. Several works consider subclasses of quantum states that can be learned in polynomial sample complexity such as stabilizer states or high-temperature Gibbs states. Other works consider a weaker sense of learning, such as PAC learning and shadow tomography. In this work, we consider learning states that are close to neural network quantum states, which can efficiently be represented by a graphical model called restricted Boltzmann machines (RBMs). To this end, we exhibit robustness results for efficient provable two-hop neighborhood learning algorithms for ferromagnetic and locally consistent RBMs. We consider the $L_p$-norm as a measure of closeness, including both total variation distance and max-norm distance in the limit. Our results allow certain quantum states to be learned with a sample complexity \textit{exponentially} better than naive tomography. We hence provide new classes of efficiently learnable quantum states and apply new strategies to learn them.

quant-ph↗

arXiv2023

MomentDiff: Generative Video Moment Retrieval from Random to Real

Video moment retrieval pursues an efficient and generalized solution to identify the specific temporal segments within an untrimmed video that correspond to a given language description. To achieve this goal, we provide a generative diffusion-based framework called MomentDiff, which simulates a typical human retrieval process from random browsing to gradual localization. Specifically, we first diffuse the real span to random noise, and learn to denoise the random noise to the original span with the guidance of similarity between text and video. This allows the model to learn a mapping from arbitrary random locations to real moments, enabling the ability to locate segments from random initialization. Once trained, MomentDiff could sample random temporal segments as initial guesses and iteratively refine them to generate an accurate temporal boundary. Different from discriminative works (e.g., based on learnable proposals or queries), MomentDiff with random initialized spans could resist the temporal location biases from datasets. To evaluate the influence of the temporal location biases, we propose two anti-bias datasets with location distribution shifts, named Charades-STA-Len and Charades-STA-Mom. The experimental results demonstrate that our efficient framework consistently outperforms state-of-the-art methods on three public benchmarks, and exhibits better generalization and robustness on the proposed anti-bias datasets. The code, model, and anti-bias evaluation datasets are available at https://github.com/IMCCretrieval/MomentDiff.

cs.CV↗

arXiv2022

Quantum algorithm for Markov Random Fields structure learning by information theoretic properties

Probabilistic graphical models play a crucial role in machine learning and have wide applications in various fields. One pivotal subset is undirected graphical models, also known as Markov random fields. In this work, we investigate the structure learning methods of Markov random fields on quantum computers. We propose a quantum algorithm for structure learning of an $r$-wise Markov Random Field with a bounded degree underlying graph, based on a nearly optimal classical greedy algorithm. The quantum algorithm provides a polynomial speed-up over the classical counterpart in terms of the number of variables. Our work demonstrates the potential merits of quantum computation over classical computation in solving some problems in machine learning.

quant-ph↗

arXiv2021

Quantum algorithm for structure learning of Markov Random Fields

Markov random fields (MRFs) appear in many problems in machine learning and statistics. From a computational learning theory point of view, a natural problem of learning MRFs arises: given samples from an MRF from a restricted class, learn the structure of the MRF, that is the neighbors of each node of the underlying graph. In this work, we start at a known near-optimal classical algorithm for this learning problem and develop a modified classical algorithm. This classical algorithm retains the run time and guarantee of the previous algorithm and enables the use of quantum subroutines. Adapting a previous quantum algorithm, the Quantum Sparsitron, we provide a polynomial quantum speedup in terms of the number of variables for learning the structure of an MRF, if the MRF has bounded degree.

quant-ph↗

arXiv2020

Quantum-classical algorithms for skewed linear systems with optimized Hadamard test

The solving of linear systems provides a rich area to investigate the use of nearer-term, noisy, intermediate-scale quantum computers. In this work, we discuss hybrid quantum-classical algorithms for skewed linear systems for over-determined and under-determined cases. Our input model is such that the columns or rows of the matrix defining the linear system are given via quantum circuits of poly-logarithmic depth and the number of circuits is much smaller than their Hilbert space dimension. Our algorithms have poly-logarithmic dependence on the dimension and polynomial dependence in other natural quantities. In addition, we present an algorithm for the special case of a factorized linear system with run time poly-logarithmic in the respective dimensions. At the core of these algorithms is the Hadamard test and in the second part of this paper we consider the optimization of the circuit depth of this test. Given an $n$-qubit and $d$-depth quantum circuit $\mathcal{C}$, we can approximate $\langle 0|\mathcal{C}|0\rangle$ using $(n + s)$ qubits and $O\left(\log s + d\log (n/s) + d\right)$-depth quantum circuits, where $s\leq n$. In comparison, the standard implementation requires $n+1$ qubits and $O(dn)$ depth. Lattice geometries underlie recent quantum supremacy experiments with superconducting devices. We also optimize the Hadamard test for an $(l_1\times l_2)$ lattice with $l_1 \times l_2 = n$, and can approximate $\langle 0|\mathcal{C} |0\rangle$ with $(n + 1)$ qubits and $O\left(d \left(l_1 + l_2\right)\right)$-depth circuits. In comparison, the standard depth is $O\left(d n^2\right)$ in this setting. Both of our optimization methods are asymptotically tight in the case of one-depth quantum circuits $\mathcal{C}$.

quant-ph↗

arXiv2020

Weakly Supervised Learning with Side Information for Noisy Labeled Images

In many real-world datasets, like WebVision, the performance of DNN based classifier is often limited by the noisy labeled data. To tackle this problem, some image related side information, such as captions and tags, often reveal underlying relationships across images. In this paper, we present an efficient weakly supervised learning by using a Side Information Network (SINet), which aims to effectively carry out a large scale classification with severely noisy labels. The proposed SINet consists of a visual prototype module and a noise weighting module. The visual prototype module is designed to generate a compact representation for each category by introducing the side information. The noise weighting module aims to estimate the correctness of each noisy image and produce a confidence score for image ranking during the training procedure. The propsed SINet can largely alleviate the negative impact of noisy image labels, and is beneficial to train a high performance CNN based classifier. Besides, we released a fine-grained product dataset called AliProducts, which contains more than 2.5 million noisy web images crawled from the internet by using queries generated from 50,000 fine-grained semantic classes. Extensive experiments on several popular benchmarks (i.e. Webvision, ImageNet and Clothing-1M) and our proposed AliProducts achieve state-of-the-art performance. The SINet has won the first place in the classification task on WebVision Challenge 2019, and outperformed other competitors by a large margin.

cs.CV↗

arXiv2020

Derivation of film theory coupled Spiegler-Kedem-Katchalsky model for membrane separation and purification

In this article, we will derive the differential equations considering the concentration polarization (\textsc{cp}) in membrane separations and purifications, which involve convective-diffusive equations of the concentration polarization layer and the Spiegler-Kedem-Katchalsky (\textsc{skk}) model of the membrane layer. Analytical solutions of these differential equations are straightforward and given. The film theory coupled \textsc{skk} model can be used to evaluate the degree of concentration polarization, and predict flux and rejection rates.

math.AP↗

arXiv2019

Bayesian inference based process design and uncertainty analysis of simulated moving bed chromatographic systems

Prominent features of simulated moving bed (SMB) chromatography processes in the downstream processing is based on the determination of operating conditions. However, effects of different types of uncertainties have to be studied and analysed whenever the triangle theory or numerical optimization approaches are applied. In this study, a Bayesian inference based method is introduced to consider the uncertainty of operating conditions on the performance assessment, of a glucose-fructose SMB unit under linear condition. A multiple chain Markov Chain Monte Carlo (MCMC) algorithm (i.e., Metropolis algorithm with delayed rejection and adjusted Metropolis) is applied to generate samples. The proposed method renders versatile information by constructing from the MCMC samples, e.g., posterior distributions, uncertainties, credible intervals of the operating conditions, and posterior predictive check and Pareto fronts between each pair of the performance indicators. Additionally, the MCMC samples can be mapped onto the $(m_\text{II}, m_\text{III})$ and $(m_\text{IV}, m_\text{I})$ plains to show the actually complete separation region under uncertainties. The proposed method is a convenient tool to find both optimal values and uncertainties of the operating conditions. Moreover, it is not limited to SMB processes under the linear isotherm; and it should be more powerful in the nonlinear scenarios.

cs.CE↗

arXiv2019

Model-based process design of a ternary protein separation using multi-step gradient ion-exchange SMB chromatography

Model-based process design of ion-exchange simulated moving bed (IEX-SMB) chromatography for center-cut separation of proteins is studied. Use of nonlinear binding models that describe more accurate adsorption behaviours of macro-molecules could make it impossible to utilize triangle theory to obtain operating parameters. Moreover, triangle theory provides no rules to design salt profiles in IEX-SMB. In the modelling study here, proteins (i.e., ribonuclease, cytochrome and lysozyme) on the chromatographic columns packed with strong cation-exchanger SP Sepharose FF is used as an example system. The general rate model with steric mass-action kinetics was used; two closed-loop IEX-SMB network schemes were investigated (i.e., cascade and eight-zone schemes). Performance of the IEX-SMB schemes was examined with respect to multi-objective indicators (i.e., purity and yield) and productivity, and compared to a single column batch system with the same amount of resin utilized. A multi-objective sampling algorithm, Markov Chain Monte Carlo (MCMC), was used to generate samples for constructing the Pareto optimal fronts. MCMC serves on the sampling purpose, which is interested in sampling the Pareto optimal points as well as those near Pareto optimal. Pareto fronts of the three schemes provide the full information of trade-off between the conflicting indicators of purity and yield. The results indicate the cascade IEX-SMB scheme and the integrated eight-zone IEX-SMB scheme have the similar performance that both outperforms the single column batch system.

cs.CE↗

arXiv2019

Compiling basic linear algebra subroutines for quantum computers

Efficiently processing basic linear algebra subroutines is of great importance for a wide range of computational problems. In this paper, we consider techniques to implement matrix functions on a quantum computer, which are composed of basic matrix operations on a set of matrices. These matrix operations include addition, multiplication, Kronecker sum, tensor product, Hadamard product, and single-matrix functions. We discuss the composed matrix functions in terms of the estimation of scalar quantities such as inner products, trace, determinant and Schatten p-norms. We thus provide a framework for compiling instructions for linear algebraic computations into gate sequences on actual quantum computers.

quant-ph↗