arXiv subjects

Ming Lu

Publications and source records attributed to Ming Lu.

At least 73 records · Page 4Linked to original sources

arXiv2025

Neural B-frame Video Compression with Bi-directional Reference Harmonization

Neural video compression (NVC) has made significant progress in recent years, while neural B-frame video compression (NBVC) remains underexplored compared to P-frame compression. NBVC can adopt bi-directional reference frames for better compression performance. However, NBVC's hierarchical coding may complicate continuous temporal prediction, especially at some hierarchical levels with a large frame span, which could cause the contribution of the two reference frames to be unbalanced. To optimize reference information utilization, we propose a novel NBVC method, termed Bi-directional Reference Harmonization Video Compression (BRHVC), with the proposed Bi-directional Motion Converge (BMC) and Bi-directional Contextual Fusion (BCF). BMC converges multiple optical flows in motion compression, leading to more accurate motion compensation on a larger scale. Then BCF explicitly models the weights of reference contexts under the guidance of motion compensation accuracy. With more efficient motions and contexts, BRHVC can effectively harmonize bi-directional references. Experimental results indicate that our BRHVC outperforms previous state-of-the-art NVC methods, even surpassing the traditional coding, VTM-RA (under random access configuration), on the HEVC datasets. The source code is released at https://github.com/kwai/NVC.

cs.CV↗

arXiv2025

StreamKV: Streaming Video Question-Answering with Segment-based KV Cache Retrieval and Compression

Video Large Language Models (Video-LLMs) have demonstrated significant potential in the areas of video captioning, search, and summarization. However, current Video-LLMs still face challenges with long real-world videos. Recent methods have introduced a retrieval mechanism that retrieves query-relevant KV caches for question answering, enhancing the efficiency and accuracy of long real-world videos. However, the compression and retrieval of KV caches are still not fully explored. In this paper, we propose \textbf{StreamKV}, a training-free framework that seamlessly equips Video-LLMs with advanced KV cache retrieval and compression. Compared to previous methods that used uniform partitioning, StreamKV dynamically partitions video streams into semantic segments, which better preserves semantic information. For KV cache retrieval, StreamKV calculates a summary vector for each segment to retain segment-level information essential for retrieval. For KV cache compression, StreamKV introduces a guidance prompt designed to capture the key semantic elements within each segment, ensuring only the most informative KV caches are retained for answering questions. Moreover, StreamKV unifies KV cache retrieval and compression within a single module, performing both in a layer-adaptive manner, thereby further improving the effectiveness of streaming video question answering. Extensive experiments on public StreamingVQA benchmarks demonstrate that StreamKV significantly outperforms existing Online Video-LLMs, achieving superior accuracy while substantially improving both memory efficiency and computational latency. The code has been released at https://github.com/sou1p0wer/StreamKV.

cs.CV↗

arXiv2025

TimeSearch-R: Adaptive Temporal Search for Long-Form Video Understanding via Self-Verification Reinforcement Learning

Temporal search aims to identify a minimal set of relevant frames from tens of thousands based on a given query, serving as a foundation for accurate long-form video understanding. Existing works attempt to progressively narrow the search space. However, these approaches typically rely on a hand-crafted search process, lacking end-to-end optimization for learning optimal search strategies. In this paper, we propose TimeSearch-R, which reformulates temporal search as interleaved text-video thinking, seamlessly integrating searching video clips into the reasoning process through reinforcement learning (RL). However, applying RL training methods, such as Group Relative Policy Optimization (GRPO), to video reasoning can result in unsupervised intermediate search decisions. This leads to insufficient exploration of the video content and inconsistent logical reasoning. To address these issues, we introduce GRPO with Completeness Self-Verification (GRPO-CSV), which gathers searched video frames from the interleaved reasoning process and utilizes the same policy model to verify the adequacy of searched frames, thereby improving the completeness of video reasoning. Additionally, we construct datasets specifically designed for the SFT cold-start and RL training of GRPO-CSV, filtering out samples with weak temporal dependencies to enhance task difficulty and improve temporal search capabilities. Extensive experiments demonstrate that TimeSearch-R achieves significant improvements on temporal search benchmarks such as Haystack-LVBench and Haystack-Ego4D, as well as long-form video understanding benchmarks like VideoMME and MLVU. Notably, TimeSearch-R establishes a new state-of-the-art on LongVideoBench with 4.1% improvement over the base model Qwen2.5-VL and 2.0% over the advanced video reasoning model Video-R1. Our code is available at https://github.com/Time-Search/TimeSearch-R.

cs.CV↗

arXiv2025

Braid group action and quasi-split affine $\imath$quantum groups I

This is the first of two papers on quasi-split affine quantum symmetric pairs $\big(\widetilde{\mathbf U}(\widehat{\mathfrak g}), \widetilde{\mathbf U}^\imath \big)$, focusing on the real rank one case, i.e., $\mathfrak{g}= \mathfrak{sl}_3$ equipped with a diagram involution. We construct explicitly a relative braid group action of type $A_2^{(2)}$ on the affine $\imath$quantum group $\widetilde{\mathbf U}^\imath$. Real and imaginary root vectors for $\widetilde{\mathbf U}^\imath$ are constructed, and a Drinfeld type presentation of $\widetilde{\mathbf U}^\imath$ is then established. This provides a new basic ingredient for the Drinfeld type presentation of higher rank quasi-split affine $\imath$quantum groups in the sequel.

math.RT↗

arXiv2025

Braid group action and quasi-split affine iquantum groups III

This is the last of three papers on Drinfeld presentations of quasi-split affine iquantum groups $\widetilde{\mathbf U}^\imath$, settling the remaining type ${\rm AIII}^{(τ)}_{2r}$. This type distinguishes itself among all quasi-split affine types in having 3 relative root lengths. Various basic real and imaginary $v$-root vectors for $\widetilde{\mathbf U}^\imath$ are constructed, giving rise to affine rank one subalgebras of $\widetilde{\mathbf U}^\imath$ associated with simple roots in the finite relative root system. We establish the relations among these $v$-root vectors and show that they provide a Drinfeld presentation of $\widetilde{\mathbf U}^\imath$.

math.RT↗

arXiv2025

Jarvis: Towards Personalized AI Assistant via Personal KV-Cache Retrieval

The rapid development of Vision-language models (VLMs) enables open-ended perception and reasoning. Recent works have started to investigate how to adapt general-purpose VLMs into personalized assistants. Even commercial models such as ChatGPT now support model personalization by incorporating user-specific information. However, existing methods either learn a set of concept tokens or train a VLM to utilize user-specific information. However, both pipelines struggle to generate accurate answers as personalized assistants. We introduce Jarvis, an innovative framework for a personalized AI assistant through personal KV-Cache retrieval, which stores user-specific information in the KV-Caches of both textual and visual tokens. The textual tokens are created by summarizing user information into metadata, while the visual tokens are produced by extracting distinct image patches from the user's images. When answering a question, Jarvis first retrieves related KV-Caches from personal storage and uses them to ensure accuracy in responses. We also introduce a fine-grained benchmark built with the same distinct image patch mining pipeline, emphasizing accurate question answering based on fine-grained user-specific information. Jarvis is capable of providing more accurate responses, particularly when they depend on specific local details. Jarvis achieves state-of-the-art results in both visual question answering and text-only tasks across multiple datasets, indicating a practical path toward personalized AI assistants. The code and dataset will be released.

cs.AI↗

arXiv2025

MixedGaussianAvatar: Realistically and Geometrically Accurate Head Avatar via Mixed 2D-3D Gaussians

Reconstructing high-fidelity 3D head avatars is crucial in various applications such as virtual reality. The pioneering methods reconstruct realistic head avatars with Neural Radiance Fields (NeRF), which have been limited by training and rendering speed. Recent methods based on 3D Gaussian Splatting (3DGS) significantly improve the efficiency of training and rendering. However, the surface inconsistency of 3DGS results in subpar geometric accuracy; later, 2DGS uses 2D surfels to enhance geometric accuracy at the expense of rendering fidelity. To leverage the benefits of both 2DGS and 3DGS, we propose a novel method named MixedGaussianAvatar for realistically and geometrically accurate head avatar reconstruction. Our main idea is to utilize 2D Gaussians to reconstruct the surface of the 3D head, ensuring geometric accuracy. We attach the 2D Gaussians to the triangular mesh of the FLAME model and connect additional 3D Gaussians to those 2D Gaussians where the rendering quality of 2DGS is inadequate, creating a mixed 2D-3D Gaussian representation. These 2D-3D Gaussians can then be animated using FLAME parameters. We further introduce a progressive training strategy that first trains the 2D Gaussians and then fine-tunes the mixed 2D-3D Gaussians. We use a unified mixed Gaussian representation to integrate the two modalities of 2D image and 3D mesh. Furthermore, the comprehensive experiments demonstrate the superiority of MixedGaussianAvatar. The code will be released.

cs.CV↗

arXiv2025

UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens

Personalized models have demonstrated remarkable success in understanding and generating concepts provided by users. However, existing methods use separate concept tokens for understanding and generation, treating these tasks in isolation. This may result in limitations for generating images with complex prompts. For example, given the concept $\langle bo\rangle$, generating "$\langle bo\rangle$ wearing its hat" without additional textual descriptions of its hat. We call this kind of generation \textit{\textbf{personalized attribute-reasoning generation}}. To address the limitation, we present UniCTokens, a novel framework that effectively integrates personalized information into a unified vision language model (VLM) for understanding and generation. UniCTokens trains a set of unified concept tokens to leverage complementary semantics, boosting two personalized tasks. Moreover, we propose a progressive training strategy with three stages: understanding warm-up, bootstrapping generation from understanding, and deepening understanding from generation to enhance mutual benefits between both tasks. To quantitatively evaluate the unified VLM personalization, we present UnifyBench, the first benchmark for assessing concept understanding, concept generation, and attribute-reasoning generation. Experimental results on UnifyBench indicate that UniCTokens shows competitive performance compared to leading methods in concept understanding, concept generation, and achieving state-of-the-art results in personalized attribute-reasoning generation. Our research demonstrates that enhanced understanding improves generation, and the generation process can yield valuable insights into understanding. Our code and dataset will be released at: \href{https://github.com/arctanxarc/UniCTokens}{https://github.com/arctanxarc/UniCTokens}.

cs.CV↗

arXiv2025

2DGS-R: Revisiting the Normal Consistency Regularization in 2D Gaussian Splatting

Recent advancements in 3D Gaussian Splatting (3DGS) have greatly influenced neural fields, as it enables high-fidelity rendering with impressive visual quality. However, 3DGS has difficulty accurately representing surfaces. In contrast, 2DGS transforms the 3D volume into a collection of 2D planar Gaussian disks. Despite advancements in geometric fidelity, rendering quality remains compromised, highlighting the challenge of achieving both high-quality rendering and precise geometric structures. This indicates that optimizing both geometric and rendering quality in a single training stage is currently unfeasible. To overcome this limitation, we present 2DGS-R, a new method that uses a hierarchical training approach to improve rendering quality while maintaining geometric accuracy. 2DGS-R first trains the original 2D Gaussians with the normal consistency regularization. Then 2DGS-R selects the 2D Gaussians with inadequate rendering quality and applies a novel in-place cloning operation to enhance the 2D Gaussians. Finally, we fine-tune the 2DGS-R model with opacity frozen. Experimental results show that compared to the original 2DGS, our method requires only 1\% more storage and minimal additional training time. Despite this negligible overhead, it achieves high-quality rendering results while preserving fine geometric structures. These findings indicate that our approach effectively balances efficiency with performance, leading to improvements in both visual fidelity and geometric reconstruction accuracy.

cs.CV↗

arXiv2025

Fast-RF-Shimming: Accelerate RF Shimming in 7T MRI using Deep Learning

Ultrahigh field (UHF) Magnetic Resonance Imaging (MRI) offers an elevated signal-to-noise ratio (SNR), enabling exceptionally high spatial resolution that benefits both clinical diagnostics and advanced research. However, the jump to higher fields introduces complications, particularly transmit radiofrequency (RF) field ($B_{1}^{+}$) inhomogeneities, manifesting as uneven flip angles and image intensity irregularities. These artifacts can degrade image quality and impede broader clinical adoption. Traditional RF shimming methods, such as Magnitude Least Squares (MLS) optimization, effectively mitigate $B_{1}^{+}$ inhomogeneity, but remain time-consuming. Recent machine learning approaches, including RF Shim Prediction by Iteratively Projected Ridge Regression and other deep learning architectures, suggest alternative pathways. Although these approaches show promise, challenges such as extensive training periods, limited network complexity, and practical data requirements persist. In this paper, we introduce a holistic learning-based framework called Fast-RF-Shimming, which achieves a 5000x speed-up compared to the traditional MLS method. In the initial phase, we employ random-initialized Adaptive Moment Estimation (Adam) to derive the desired reference shimming weights from multi-channel $B_{1}^{+}$ fields. Next, we train a Residual Network (ResNet) to map $B_{1}^{+}$ fields directly to the ultimate RF shimming outputs, incorporating the confidence parameter into its loss function. Finally, we design Non-uniformity Field Detector (NFD), an optional post-processing step, to ensure the extreme non-uniform outcomes are identified. Comparative evaluations with standard MLS optimization underscore notable gains in both processing speed and predictive accuracy, which indicates that our technique shows a promising solution for addressing persistent inhomogeneity challenges.

cs.CV↗

arXiv2025

Right-eigenstate-based approach to non-Hermitian superfluidity with two-body loss

We theoretically explore a non-Hermitian superfluid model with complex-valued interaction, inspired by two-body loss stemming from inelastic scattering observed in ultracold atomic experiments. Utilizing both the right-eigenstate-based mean-field theory and its biorthogonal counterpart, we study the properties of the system. Notably, the right-eigenstate-based framework produces smooth and continuous solutions, in stark contrast to the absence of nontrivial solutions and the abrupt discontinuities observed in the biorthogonal-eigenstate-based framework under moderate dissipation. In addition, the lower condensation energy obtained in the former framework suggests its superior suitability for describing this system. Furthermore, we explore the impact of backscattering, a crucial factor in realistic systems. Our analysis reveals that, facilitated by two-body loss, even moderate backscattering destabilizes the superfluid state. Sufficiently strong backscattering completely destroys it, highlighting a key mechanism for the fragility of this non-Hermitian quantum phase.

cond-mat.quant-gas↗

arXiv2025

BranchGRPO: Stable and Efficient GRPO with Structured Branching in Diffusion Models

Recent progress in aligning image and video generative models with Group Relative Policy Optimization (GRPO) has improved human preference alignment, but existing variants remain inefficient due to sequential rollouts and large numbers of sampling steps, unreliable credit assignment: sparse terminal rewards are uniformly propagated across timesteps, failing to capture the varying criticality of decisions during denoising. In this paper, we present BranchGRPO, a method that restructures the rollout process into a branching tree, where shared prefixes amortize computation and pruning removes low-value paths and redundant depths. BranchGRPO introduces three contributions: (1) a branching scheme that amortizes rollout cost through shared prefixes while preserving exploration diversity; (2) a reward fusion and depth-wise advantage estimator that transforms sparse terminal rewards into dense step-level signals; and (3) pruning strategies that cut gradient computation but leave forward rollouts and exploration unaffected. On HPDv2.1 image alignment, BranchGRPO improves alignment scores by up to \textbf{16\%} over DanceGRPO, while reducing per-iteration training time by nearly \textbf{55\%}. A hybrid variant, BranchGRPO-Mix, further accelerates training to 4.7x faster than DanceGRPO without degrading alignment. On WanX video generation, it further achieves higher Video-Align scores with sharper and temporally consistent frames compared to DanceGRPO. Codes are available at \href{https://fredreic1849.github.io/BranchGRPO-Webpage/}{BranchGRPO}.

cs.CV↗

arXiv2025

Preparation of the First Cu-based Nb$_3$Sn Sample via Bronze Route for Quadrupole Resonator Testing

We report the first successful production of a Cu-based Nb$_3$Sn sample specifically designed for Quadrupole Resonator (QPR) testing, representing a significant step toward scalable RF superconducting coatings of Nb$_3$Sn on copper substrates. The sample was fabricated using an optimized electrochemical thermal synthesis (ETS) via the bronze route, incorporating several key advancements: electropolishing of the Cu substrate, electroplating of the bronze precursor layer, a tailored heat treatment at approximately 700 $^\circ$C to promote grain growth and suppress tin-rich impurity phases, and a newly developed chemical etching procedure for effective removal of surface bronze residues and contaminants. These improvements address longstanding challenges in the fabrication of high-quality Cu-based Nb$_3$Sn thin films. Subsequent QPR measurements yielded the peak magnetic field and temperature dependent surface resistance $R_s$, as well as the superconducting transition temperature and quench field. Although the achieved RF performance -- characterized by a minimum $R_s$ of 43.4 n$Ω$ at 4.5 K and 15 mT -- is not yet optimal, the results clearly demonstrate the feasibility of this approach and its potential for further enhancement through process refinement.

physics.app-ph↗

arXiv2025

Low-Cost and Detunable Wireless Resonator Glasses for Enhanced Eye MRI with Concurrent High-Quality Whole Brain MRI

Purpose: To develop and evaluate a wearable wireless resonator glasses design that enhances eye MRI signal-to-noise ratio (SNR) without compromising whole-brain image quality at 7 T. Methods: The device integrates two detunable LC loop resonators into a lightweight, 3D-printed frame positioned near the eyes. The resonators passively couple to a standard 2Tx/32Rx head coil without hardware modifications. Bench tests assessed tuning, isolation, and detuning performance. B1$^+$ maps were measured in a head/shoulder phantom, and SNR maps were obtained in both phantom and in vivo experiments. Results: Bench measurements confirmed accurate tuning, strong inter-element isolation, and effective passive detuning. Phantom B1$^+$ mapping showed negligible differences between configurations with and without the resonators. Phantom and in vivo imaging demonstrated up to about a 3-fold SNR gain in the eye region, with no measurable SNR loss in the brain. Conclusion: The wireless resonator glasses provide a low-cost, easy-to-use solution that improves ocular SNR while preserving whole-brain image quality, enabling both dedicated eye MRI and simultaneous eye-brain imaging at ultrahigh field.

eess.IV↗

arXiv2025

First-Order PT Phase Transition in Non-Hermitian Superconductors

The interplay between superconductivity and environmental dissipation, effectively captured by non-Hermitian Hamiltonian, is a new frontier for exotic quantum phases. We explore a PT-symmetric non-Hermitian superconductor with balanced gain and loss. To ensure experimental relevance, we develop a right-eigenstate-based non-Hermitian mean-field theory. We uncover a novel first-order phase transition that coincides exactly with the PT symmetry breaking point, driven by the interplay between superconducting pairing interactions and non-Hermitian dissipation. In the PT-symmetric phase, moderate NH dissipation enhances superconductivity, while in the PT-broken phase, intensified dissipation significantly suppresses it. These phenomena, characterized by abrupt jumps in observables, can be probed through localized spectral measurements and macroscopic superfluid density analysis. Additionally, the stability analysis offers robust theoretical insights to support experimental investigations of this unique transition.

cond-mat.supr-con↗

arXiv2025

MMG-Vid: Maximizing Marginal Gains at Segment-level and Token-level for Efficient Video LLMs

Video Large Language Models (VLLMs) excel in video understanding, but their excessive visual tokens pose a significant computational challenge for real-world applications. Current methods aim to enhance inference efficiency by visual token pruning. However, they do not consider the dynamic characteristics and temporal dependencies of video frames, as they perceive video understanding as a multi-frame task. To address these challenges, we propose MMG-Vid, a novel training-free visual token pruning framework that removes redundancy by Maximizing Marginal Gains at both segment-level and token-level. Specifically, we first divide the video into segments based on frame similarity, and then dynamically allocate the token budget for each segment to maximize the marginal gain of each segment. Subsequently, we propose a temporal-guided DPC algorithm that jointly models inter-frame uniqueness and intra-frame diversity, thereby maximizing the marginal gain of each token. By combining both stages, MMG-Vid can maximize the utilization of the limited token budget, significantly improving efficiency while maintaining strong performance. Extensive experiments demonstrate that MMG-Vid can maintain over 99.5% of the original performance, while effectively reducing 75% visual tokens and accelerating the prefilling stage by 3.9x on LLaVA-OneVision-7B. Code will be released soon.

cs.CV↗

arXiv2025

Mn4Al11: A Half-Semimetal Candidate with Anomalous Electronic Behaviors

Half-semimetals, characterized by their spin-polarized electronic states, hold significant promise for spintronic applications but remain scarce due to stringent electronic and magnetic criteria. Through a combination of transport measurements and optical spectroscopy, we investigated the intermetallic compound Mn4Al11, which features an exceptionally low carrier concentration and undergoes a magnetic phase transition near 68 K. Transport measurements reveal anomalies that deviate from typical metallic behavior at low temperatures. Optical spectroscopy indicates a small, nearly frequency-independent optical conductivity in the far-infrared region, with spectral weight decreasing as the temperature drops from 300 K to 50 K. These behaviors suggest a temperaturedependent carrier density and significant scattering of charge carriers. Combining experimental findings with calculated electronic band structures, we propose that Mn4Al11 is a novel half-semimetal candidate exhibiting a ferrimagnetic ground state.

cond-mat.str-el↗

arXiv2025

Small-Large Collaboration: Training-efficient Concept Personalization for Large VLM using a Meta Personalized Small VLM

Personalizing Vision-Language Models (VLMs) to transform them into daily assistants has emerged as a trending research direction. However, leading companies like OpenAI continue to increase model size and develop complex designs such as the chain of thought (CoT). While large VLMs are proficient in complex multi-modal understanding, their high training costs and limited access via paid APIs restrict direct personalization. Conversely, small VLMs are easily personalized and freely available, but they lack sufficient reasoning capabilities. Inspired by this, we propose a novel collaborative framework named Small-Large Collaboration (SLC) for large VLM personalization, where the small VLM is responsible for generating personalized information, while the large model integrates this personalized information to deliver accurate responses. To effectively incorporate personalized information, we develop a test-time reflection strategy, preventing the potential hallucination of the small VLM. Since SLC only needs to train a meta personalized small VLM for the large VLMs, the overall process is training-efficient. To the best of our knowledge, this is the first training-efficient framework that supports both open-source and closed-source large VLMs, enabling broader real-world personalized applications. We conduct thorough experiments across various benchmarks and large VLMs to demonstrate the effectiveness of the proposed SLC framework. The code will be released at https://github.com/Hhankyangg/SLC.

cs.CV↗