Searcharxiv⌕ Search

arXiv subjects

Dong Zhang

Publications and source records attributed to Dong Zhang.

At least 55 records · Page 3Linked to original sources
arXiv2026

Training Beyond Convergence: Grokking nnU-Net for Glioma Segmentation in Sub-Saharan MRI

Gliomas are placing an increasingly clinical burden on Sub-Saharan Africa (SSA). In the region, the median survival for patients remains under two years, and access to diagnostic imaging is extremely limited. These constraints highlight an urgent need for automated tools that can extract the maximum possible information from each available scan, tools that are specifically trained on local data, rather than adapted from high-income settings where conditions are vastly different. We utilize the Brain Tumor Segmentation (BraTS) Africa 2025 Challenge dataset, an expert annotated collection of glioma MRIs. Our objectives are: (i) establish a strong baseline with nnUNet on this dataset, and (ii) explore whether the celebrated "grokking" phenomenon an abrupt, late training jump from memorization to superior generalization can be triggered to push performance without extra labels. We evaluate two training regimes. The first is a fast, budget-conscious approach that limits optimization to just a few epochs, reflecting the constrained GPU resources typically available in African institutions. Despite this limitation, nnUNet achieves strong Dice scores: 92.3% for whole tumor (WH), 86.6% for tumor core (TC), and 86.3% for enhancing tumor (ET). The second regime extends training well beyond the point of convergence, aiming to trigger a grokking-driven performance leap. With this approach, we were able to achieve grokking and enhanced our results to higher Dice scores: 92.2% for whole tumor (WH), 90.1% for tumor core (TC), and 90.2% for enhancing tumor (ET).

eess.IV↗
arXiv2026

Agentic AI for Self-Driving Laboratories in Soft Matter: Taxonomy, Benchmarks,and Open Challenges

Self-driving laboratories (SDLs) close the loop between experiment design, automated execution, and data-driven decision making, and they provide a demanding testbed for agentic AI under expensive actions, noisy and delayed feedback, strict feasibility and safety constraints, and non-stationarity. This survey uses soft matter as a representative setting but focuses on the AI questions that arise in real laboratories. We frame SDL autonomy as an agent environment interaction problem with explicit observations, actions, costs, and constraints, and we use this formulation to connect common SDL pipelines to established AI principles. We review the main method families that enable closed loop experimentation, including Bayesian optimization and active learning for sample efficient experiment selection, planning and reinforcement learning for long horizon protocol optimization, and tool using agents that orchestrate heterogeneous instruments and software. We emphasize verifiable and provenance aware policies that support debugging, reproducibility, and safe operation. We then propose a capability driven taxonomy that organizes systems by decision horizon, uncertainty modeling, action parameterization, constraint handling, failure recovery, and human involvement. To enable meaningful comparison, we synthesize benchmark task templates and evaluation metrics that prioritize cost aware performance, robustness to drift, constraint violation behavior, and reproducibility. Finally, we distill lessons from deployed SDLs and outline open challenges in multi-modal representation, calibrated uncertainty, safe exploration, and shared benchmark infrastructure.

cs.AI↗
arXiv2026

Equivalent spectral theory for fundamental graph cut problems

We introduce and develop equivalent spectral graph theory for several fundamental graph cut problems including maxcut, mincut, Cheeger cut, anti-Cheeger cut, dual Cheeger problem and their useful variants. A specified strategy for achieving an equivalent eigenproblem is proposed for a general graph cut problem via the set-pair Lovász extension and the Dinkelbach scheme. For a class of 2-cut and 3-cut problems, we reveal the intrinsic difference-of-submodularity for the fractional formulations and show that their set-pair Lovász extensions yield equivalent difference-of-convex structures. Building on the Dinkelbach scheme, we finally establish a unified research roadmap for nonlinear spectral theory that provides a one-to-one correspondence between certain eigenpairs and the optimal graph cut problems. The finer structure of the eigenvectors, the Courant nodal domain theorem and the graphic feature of eigenvalues are studied systematically in the setting of these new nonlinear eigenproblems.

math.CO↗
arXiv2026

Bridging Semantic Understanding and Popularity Bias with LLMs

Semantic understanding of popularity bias is a crucial yet underexplored challenge in recommender systems, where popular items are often favored at the expense of niche content. Most existing debiasing methods treat the semantic understanding of popularity bias as a matter of diversity enhancement or long-tail coverage, neglecting the deeper semantic layer that embodies the causal origins of the bias itself. Consequently, such shallow interpretations limit both their debiasing effectiveness and recommendation accuracy. In this paper, we propose FairLRM, a novel framework that bridges the gap in the semantic understanding of popularity bias with Recommendation via Large Language Model (RecLLM). FairLRM decomposes popularity bias into item-side and user-side components, using structured instruction-based prompts to enhance the model's comprehension of both global item distributions and individual user preferences. Unlike traditional methods that rely on surface-level features such as "diversity" or "debiasing", FairLRM improves the model's ability to semantically interpret and address the underlying bias. Through empirical evaluation, we show that FairLRM significantly enhances both fairness and recommendation accuracy, providing a more semantically aware and trustworthy approach to enhance the semantic understanding of popularity bias. The implementation is available at https://github.com/LuoRenqiang/FairLRM.

cs.IR↗
arXiv2026

MiMo-V2-Flash Technical Report

We present MiMo-V2-Flash, a Mixture-of-Experts (MoE) model with 309B total parameters and 15B active parameters, designed for fast, strong reasoning and agentic capabilities. MiMo-V2-Flash adopts a hybrid attention architecture that interleaves Sliding Window Attention (SWA) with global attention, with a 128-token sliding window under a 5:1 hybrid ratio. The model is pre-trained on 27 trillion tokens with Multi-Token Prediction (MTP), employing a native 32k context length and subsequently extended to 256k. To efficiently scale post-training compute, MiMo-V2-Flash introduces a novel Multi-Teacher On-Policy Distillation (MOPD) paradigm. In this framework, domain-specialized teachers (e.g., trained via large-scale reinforcement learning) provide dense and token-level reward, enabling the student model to perfectly master teacher expertise. MiMo-V2-Flash rivals top-tier open-weight models such as DeepSeek-V3.2 and Kimi-K2, despite using only 1/2 and 1/3 of their total parameters, respectively. During inference, by repurposing MTP as a draft model for speculative decoding, MiMo-V2-Flash achieves up to 3.6 acceptance length and 2.6x decoding speedup with three MTP layers. We open-source both the model weights and the three-layer MTP weights to foster open research and community collaboration.

cs.CL↗
arXiv2026

WESR: Scaling and Evaluating Word-level Event-Speech Recognition

Speech conveys not only linguistic information but also rich non-verbal vocal events such as laughing and crying. While semantic transcription is well-studied, the precise localization of non-verbal events remains a critical yet under-explored challenge. Current methods suffer from insufficient task definitions with limited category coverage and ambiguous temporal granularity. They also lack standardized evaluation frameworks, hindering the development of downstream applications. To bridge this gap, we first develop a refined taxonomy of 21 vocal events, with a new categorization into discrete (standalone) versus continuous (mixed with speech) types. Based on the refined taxonomy, we introduce WESR-Bench, an expert-annotated evaluation set (900+ utterances) with a novel position-aware protocol that disentangles ASR errors from event detection, enabling precise localization measurement for both discrete and continuous events. We also build a strong baseline by constructing a 1,700+ hour corpus, and train specialized models, surpassing both open-source audio-language models and commercial APIs while preserving ASR quality. We anticipate that WESR will serve as a foundational resource for future research in modeling rich, real-world auditory scenes.

cs.CL↗
arXiv2026

Modeling Item-Level Dynamic Variability with Residual Diffusion for Bundle Recommendation

Existing solutions for bundle recommendation (BR) have achieved remarkable effectiveness for predicting the user's preference for prebuilt bundles. However, bundle-item (B-I) affiliation will vary dynamically in real scenarios. For example, a bundle themed as 'casual outfit' may add 'hat' or remove 'watch' due to factors such as seasonal variations, changes in user preferences or inventory adjustments. Our empirical study demonstrates that the performance of mainstream BR models may fluctuate or decline under item-level variability. This paper makes the first attempt to address the above problem and proposes a novel Residual Diffusion for Bundle Recommendation(RDiffBR)asamodel-agnostic generative framework which can assist a BR model in adapting this scenario. During the initial training of the BR model, RDiffBR employs a residual diffusion model to process the item-level bundle embeddings which are generated by the BR model to represent bundle theme via a forward-reverse process. In the inference stage, RDiffBR reverses item-level bundle embeddings obtained by the well-trained bundle model under B-I variability scenarios to generate the effective item level bundle embeddings. In particular, the residual connection in our residual approximator significantly enhances BR models' ability to generate high-quality item-level bundle embeddings. Experiments on six BR models and four public datasets from different domains show that RDiffBR improves the performance of Recall and NDCG of backbone BR models by up to 23%, while only increases training time about 4%.

cs.IR↗
arXiv2025

MiMo-Audio: Audio Language Models are Few-Shot Learners

Existing audio language models typically rely on task-specific fine-tuning to accomplish particular audio tasks. In contrast, humans are able to generalize to new audio tasks with only a few examples or simple instructions. GPT-3 has shown that scaling next-token prediction pretraining enables strong generalization capabilities in text, and we believe this paradigm is equally applicable to the audio domain. By scaling MiMo-Audio's pretraining data to over one hundred million of hours, we observe the emergence of few-shot learning capabilities across a diverse set of audio tasks. We develop a systematic evaluation of these capabilities and find that MiMo-Audio-7B-Base achieves SOTA performance on both speech intelligence and audio understanding benchmarks among open-source models. Beyond standard metrics, MiMo-Audio-7B-Base generalizes to tasks absent from its training data, such as voice conversion, style transfer, and speech editing. MiMo-Audio-7B-Base also demonstrates powerful speech continuation capabilities, capable of generating highly realistic talk shows, recitations, livestreaming and debates. At the post-training stage, we curate a diverse instruction-tuning corpus and introduce thinking mechanisms into both audio understanding and generation. MiMo-Audio-7B-Instruct achieves open-source SOTA on audio understanding benchmarks (MMSU, MMAU, MMAR, MMAU-Pro), spoken dialogue benchmarks (Big Bench Audio, MultiChallenge Audio) and instruct-TTS evaluations, approaching or surpassing closed-source models. Model checkpoints and full evaluation suite are available at https://github.com/XiaomiMiMo/MiMo-Audio.

cs.CL↗
arXiv2025

AI Code in the Wild: Measuring Security Risks and Ecosystem Shifts of AI-Generated Code in Modern Software

Large language models (LLMs) for code generation are becoming integral to modern software development, but their real-world prevalence and security impact remain poorly understood. We present the first large-scale empirical study of AI-generated code (AIGCode) in the wild. We build a high-precision detection pipeline and a representative benchmark to distinguish AIGCode from human-written code, and apply them to (i) development commits from the top 1,000 GitHub repositories (2022-2025) and (ii) 7,000+ recent CVE-linked code changes. This lets us label commits, files, and functions along a human/AI axis and trace how AIGCode moves through projects and vulnerability life cycles. Our measurements show three ecological patterns. First, AIGCode is already a substantial fraction of new code, but adoption is structured: AI concentrates in glue code, tests, refactoring, documentation, and other boilerplate, while core logic and security-critical configurations remain mostly human-written. Second, adoption has security consequences: some CWE families are overrepresented in AI-tagged code, and near-identical insecure templates recur across unrelated projects, suggesting "AI-induced vulnerabilities" propagated by shared models rather than shared maintainers. Third, in human-AI edit chains, AI introduces high-throughput changes while humans act as security gatekeepers; when review is shallow, AI-introduced defects persist longer, remain exposed on network-accessible surfaces, and spread to more files and repositories. We will open-source the complete dataset and release analysis artifacts and fine-grained documentation of our methodology and findings.

cs.SE↗
arXiv2025

MMRAG-RFT: Two-stage Reinforcement Fine-tuning for Explainable Multi-modal Retrieval-augmented Generation

Multi-modal Retrieval-Augmented Generation (MMRAG) enables highly credible generation by integrating external multi-modal knowledge, thus demonstrating impressive performance in complex multi-modal scenarios. However, existing MMRAG methods fail to clarify the reasoning logic behind retrieval and response generation, which limits the explainability of the results. To address this gap, we propose to introduce reinforcement learning into multi-modal retrieval-augmented generation, enhancing the reasoning capabilities of multi-modal large language models through a two-stage reinforcement fine-tuning framework to achieve explainable multi-modal retrieval-augmented generation. Specifically, in the first stage, rule-based reinforcement fine-tuning is employed to perform coarse-grained point-wise ranking of multi-modal documents, effectively filtering out those that are significantly irrelevant. In the second stage, reasoning-based reinforcement fine-tuning is utilized to jointly optimize fine-grained list-wise ranking and answer generation, guiding multi-modal large language models to output explainable reasoning logic in the MMRAG process. Our method achieves state-of-the-art results on WebQA and MultimodalQA, two benchmark datasets for multi-modal retrieval-augmented generation, and its effectiveness is validated through comprehensive ablation experiments.

cs.AI↗
arXiv2025

Argus: A Multi-Agent Sensitive Information Leakage Detection Framework Based on Hierarchical Reference Relationships

Sensitive information leakage in code repositories has emerged as a critical security challenge. Traditional detection methods that rely on regular expressions, fingerprint features, and high-entropy calculations often suffer from high false-positive rates. This not only reduces detection efficiency but also significantly increases the manual screening burden on developers. Recent advances in large language models (LLMs) and multi-agent collaborative architectures have demonstrated remarkable potential for tackling complex tasks, offering a novel technological perspective for sensitive information detection. In response to these challenges, we propose Argus, a multi-agent collaborative framework for detecting sensitive information. Argus employs a three-tier detection mechanism that integrates key content, file context, and project reference relationships to effectively reduce false positives and enhance overall detection accuracy. To comprehensively evaluate Argus in real-world repository environments, we developed two new benchmarks, one to assess genuine leak detection capabilities and another to evaluate false-positive filtering performance. Experimental results show that Argus achieves up to 94.86% accuracy in leak detection, with a precision of 96.36%, recall of 94.64%, and an F1 score of 0.955. Moreover, the analysis of 97 real repositories incurred a total cost of only 2.2$. All code implementations and related datasets are publicly available at https://github.com/TheBinKing/Argus-Guard for further research and application.

cs.CR↗
arXiv2025

Re-examining the boundary conditions in modelling SAW-driven acoustofluidic streaming

Numerical simulations of surface acoustic wave (SAW)-induced acoustic streaming are highly sensitive to the choice of second-order boundary conditions. This study systematically compares the no-slip (NS) and Stokes slip (SD) boundary conditions through different numerical approaches. Two- and three-dimensional simulations based on the Reynolds stress method are performed for standing SAW and travelling SAW devices. Results are validated against particle image velocimetry measurements of streaming patterns and velocities. We show that the SD condition yields Lagrangian velocity fields in significantly better agreement with experiments than the NS condition, accurately capturing vortex number, rotation direction, and amplitude across varying device geometries and operating conditions. In contrast, the NS condition overpredicts velocities by 1-2 orders of magnitude and often fails to reproduce experimentally observed vortex structures. These findings highlight the essential role of the Stokes drift boundary condition in modelling acoustic streaming and provide clear guidance for its use in future simulations of SAW-based acoustofluidic systems.

physics.flu-dyn↗
arXiv2025

Interaction-Centric Knowledge Infusion and Transfer for Open-Vocabulary Scene Graph Generation

Open-vocabulary scene graph generation (OVSGG) extends traditional SGG by recognizing novel objects and relationships beyond predefined categories, leveraging the knowledge from pre-trained large-scale models. Existing OVSGG methods always adopt a two-stage pipeline: 1) \textit{Infusing knowledge} into large-scale models via pre-training on large datasets; 2) \textit{Transferring knowledge} from pre-trained models with fully annotated scene graphs during supervised fine-tuning. However, due to a lack of explicit interaction modeling, these methods struggle to distinguish between interacting and non-interacting instances of the same object category. This limitation induces critical issues in both stages of OVSGG: it generates noisy pseudo-supervision from mismatched objects during knowledge infusion, and causes ambiguous query matching during knowledge transfer. To this end, in this paper, we propose an inter\textbf{AC}tion-\textbf{C}entric end-to-end OVSGG framework (\textbf{ACC}) in an interaction-driven paradigm to minimize these mismatches. For \textit{interaction-centric knowledge infusion}, ACC employs a bidirectional interaction prompt for robust pseudo-supervision generation to enhance the model's interaction knowledge. For \textit{interaction-centric knowledge transfer}, ACC first adopts interaction-guided query selection that prioritizes pairing interacting objects to reduce interference from non-interacting ones. Then, it integrates interaction-consistent knowledge distillation to bolster robustness by pushing relational foreground away from the background while retaining general knowledge. Extensive experimental results on three benchmarks show that ACC achieves state-of-the-art performance, demonstrating the potential of interaction-centric paradigms for real-world applications.

cs.CV↗
arXiv2025

Nexus: An Omni-Perceptive And -Interactive Model for Language, Audio, And Vision

This work proposes an industry-level omni-modal large language model (LLM) pipeline that integrates auditory, visual, and linguistic modalities to overcome challenges such as limited tri-modal datasets, high computational costs, and complex feature alignments. Our pipeline consists of three main components: First, a modular framework enabling flexible configuration of various encoder-LLM-decoder architectures. Second, a lightweight training strategy that pre-trains audio-language alignment on the state-of-the-art vision-language model Qwen2.5-VL, thus avoiding the costly pre-training of vision-specific modalities. Third, an audio synthesis pipeline that generates high-quality audio-text data from diverse real-world scenarios, supporting applications such as Automatic Speech Recognition and Speech-to-Speech chat. To this end, we introduce an industry-level omni-modal LLM, Nexus. Extensive experiments validate the efficacy of our pipeline, yielding the following key findings:(1) In the visual understanding task, Nexus exhibits superior performance compared with its backbone model - Qwen2.5-VL-7B, validating the efficiency of our training strategy. (2) Within the English Spoken Question-Answering task, the model achieves better accuracy than the same-period competitor (i.e, MiniCPM-o2.6-7B) in the LLaMA Q. benchmark. (3) In our real-world ASR testset, Nexus achieves outstanding performance, indicating its robustness in real scenarios. (4) In the Speech-to-Text Translation task, our model outperforms Qwen2-Audio-Instruct-7B. (5) In the Text-to-Speech task, based on pretrained vocoder (e.g., Fishspeech1.4 or CosyVoice2.0), Nexus is comparable to its backbone vocoder on Seed-TTS benchmark. (6) An in-depth analysis of tri-modal alignment reveals that incorporating the audio modality enhances representational alignment between vision and language.

cs.MM↗
arXiv2025

Combinatorial Courant-Fischer-Weyl Minimax Principle on Cheeger $k$-constants of Weighted Forests

We establish novel max-min and minimax characterizations of Cheeger $k$-constants in weighted forests, thereby providing the first combinatorial analogue of the Courant-Fischer-Weyl minimax principle. As for applications, we prove that the forest 1-Laplacian variational eigenvalues are independent of the choice of typical indexes; we propose a refined higher order Cheeger inequality involving numbers of loops of graphs and $p$-Laplacian eigenvalues; and we present a combinatorial proof for the equality $h_k=λ_k(Δ_1)$ which connects the 1-Laplacian variational eigenvalues and the multiway Cheeger constants.

math.CO↗
arXiv2025

Finite projective planes meet spectral gaps

We show that for any connected graph $G$ with maximum degree $d\ge3$, the spectral gap from $0$ with respect to the adjacency matrix is at most $\sqrt{d-1}$, with equality if and only if $G$ is the incidence graph of a finite projective plane of order $d-1$; and for other cases, the bound $\sqrt{d-1}$ is improved to $\sqrt{d-2}$. This is a spectral gap version of a result by Mohar and Tayfeh-Rezaie. Moreover, for $d$-regular graphs with girth at least 7, the bound $\sqrt{d-2}$ is further improved to $\sqrt{d-c(d)}$ where $c(d)\ge 2$ and $\lim\limits_{d\to\infty}c(d)/d=(\sqrt{5}-1)/2$. A similar yet more subtle phenomenon involving the normalized Laplacian is also investigated, where we work on graphs of degrees $\ge d$ rather than $\le d$. We prove that for any graph $G$ with \emph{minimum} degree $d\ge 3$, the spectral gap from the value 1 with respect to the normalized Laplacian is at most $\sqrt{d-1}/d$, with equality if and only if $G$ is the incidence graph of a finite projective plane of order $d-1$. As an application, we provide a new sharp bound for the convergence rate of some eigenvalues of the Laplacian on the weighted neighborhood graphs introduced by Bauer and Jost.

math.CO↗
arXiv2025

Token Merging via Spatiotemporal Information Mining for Surgical Video Understanding

Vision Transformer models have shown impressive effectiveness in the surgical video understanding tasks through long-range dependency modeling. However, current methods suffer from prohibitive computational costs due to processing massive spatiotemporal tokens across video frames. While prior work on token merging has advanced model efficiency, they fail to adequately consider the inherent spatiotemporal structure of video data and overlook the heterogeneous nature of information distribution, leading to suboptimal performance. In this paper, we propose a spatiotemporal information mining token merging (STIM-TM) method, representing the first dedicated approach for surgical video understanding. STIM-TM introduces a decoupled strategy that reduces token redundancy along temporal and spatial dimensions independently. Specifically, the temporal component merges spatially corresponding tokens from consecutive frames using saliency weighting, preserving critical sequential information and maintaining continuity. Meanwhile, the spatial component prioritizes merging static tokens through temporal stability analysis, protecting dynamic regions containing essential surgical information. Operating in a training-free manner, STIM-TM achieves significant efficiency gains with over $65\%$ GFLOPs reduction while preserving competitive accuracy across comprehensive surgical video tasks. Our method also supports efficient training of long-sequence surgical videos, addressing computational bottlenecks in surgical applications.

cs.CV↗