Searcharxiv⌕ Search

arXiv subjects

Yue Zhao

Publications and source records attributed to Yue Zhao.

At least 109 records · Page 6Linked to original sources
arXiv2026

MeTHanol: Modularized Thinking Language Models with Intermediate Layer Thinking, Decoding and Bootstrapping Reasoning

Current research efforts are focused on enhancing the thinking and reasoning capability of large language model (LLM) by prompting, data-driven emergence and inference-time computation. In this study, we consider stimulating language model's thinking and cognitive abilities from a modular perspective, which mimics the human brain architecture. We select a specific intermediate attention layer with newly implemented language heads. We conduct dual-layer fine-tuning by annotated (query, thought, answer) samples and show that the intermediate layer can also learn to decode fluent and reasonable language tokens. A two-pass inference mechanism is designed to generate thoughts then formal responses. The entire framework is called modularized thinking language model (MeTHanol) which can enhance LLM's cognitive behaviors as indicated by Theory of Mind (ToM) and Vignette-based experiments. Case studies also show that MeTHanol can plan and self-reflect and generate human-like thoughts and answers, even on unseen and open-domain tasks. MeTHanol can also adapt to a personalized prompt and behave as the specified character. Our study holds promise for significant cognitive gains from a modular perspective. Our code, model and data are available at https://bachozean.github.io/methanol-page

cs.CL↗
arXiv2026

Ti-Audio: The First Multi-Dialectal End-to-End Speech LLM for Tibetan

Recent advances in Speech Large Language Models (Speech-LLMs) have made significant progress, greatly enhancing multimodal interaction capabilities.However, their application in low-resource and dialect-diverse environments still faces challenges. The severe scarcity of Tibetan data, coupled with the phonetic differences among its major dialects (Ü-Tsang, Amdo, and Kham), is a prime example of this challenge. This paper proposes Ti-Audio, the first multi-dialectal end-to-end Speech-LLM for Tibetan. To efficiently align speech and text, we introduce a Dynamic Q-Former Adapter that extracts essential acoustic features from variable-length speech, ensuring stable cross-modal alignment even with limited data. At the data level, we leverage mutual assistance among related dialects to alleviate data scarcity and employ a temperature-based sampling strategy to maximize this synergy. Experimental results demonstrate that Ti-Audio achieves state-of-the-art performance on Tibetan benchmarks for automatic speech recognition and speech translation. Our work validates the effectiveness of cross-dialectal cooperation and provides a scalable paradigm for the development of Speech-LLM in low-resource scenarios.

cs.SD↗
arXiv2026

A Survey on LLM-based Conversational User Simulation

User simulation has long played a vital role in computer science due to its potential to support a wide range of applications. Language, as the primary medium of human communication, forms the foundation of social interaction and behavior. Consequently, simulating conversational behavior has become a key area of study. Recent advancements in large language models (LLMs) have significantly catalyzed progress in this domain by enabling high-fidelity generation of synthetic user conversation. In this paper, we survey recent advancements in LLM-based conversational user simulation. We introduce a novel taxonomy covering user granularity and simulation objectives. Additionally, we systematically analyze core techniques and evaluation methodologies. We aim to keep the research community informed of the latest advancements in conversational user simulation and to further facilitate future research by identifying open challenges and organizing existing work under a unified framework.

cs.CL↗
arXiv2026

Formulae for the Drazin inverse of Modified Tensors via the Einstein Product

This paper establishes exact expressions for the Drazin inverse of the modified tensor $\mathcal A-\mathcal C*_N\mathcal D^D*_N\mathcal B$ via the Einstein product, formulated using the Drazin inverse of $\mathcal A$ and the generalized Schur complement $\mathcal D-\mathcal B*_N\mathcal A^{D}*_N\mathcal C$, providing a comprehensive generalization and unification of existing results in the literature for the case when the tensors are of order two. Furthermore, the findings reduce to the classical Sherman-Morrison-Woodbury formula in the special case of second-order tensors. Finally, we give an example to illustrate our new explicit expression.

math.NA↗
arXiv2026

CoAct: Co-Active LLM Preference Learning with Human-AI Synergy

Learning from preference-based feedback has become an effective approach for aligning LLMs across diverse tasks. However, high-quality human-annotated preference data remains expensive and scarce. Existing methods address this challenge through either self-rewarding, which scales by using purely AI-generated labels but risks unreliability, or active learning, which ensures quality through oracle annotation but cannot fully leverage unlabeled data. In this paper, we present CoAct, a novel framework that synergistically combines self-rewarding and active learning through strategic human-AI collaboration. CoAct leverages self-consistency to identify both reliable self-labeled data and samples that require oracle verification. Additionally, oracle feedback guides the model to generate new instructions within its solvable capability. Evaluated on three reasoning benchmarks across two model families, CoAct achieves average improvements of +13.25% on GSM8K, +8.19% on MATH, and +13.16% on WebInstruct, consistently outperforming all baselines.

cs.CL↗
arXiv2026

Weak Magnetic Sensing via Floquet Driving in an Active Cavity Magnon Coupled System

While significant advancements have been made in weak magnetic field detection, conventional high-sensitivity techniques are often limited by requirements for cryogenic operation or bulky setups. In this work, we develop a sensitive alternating magnetic field sensor based on a coupled system of an active microwave cavity and yttrium iron garnet (YIG), with the components implemented on printed circuit boards (PCBs). By introducing electrically tunable gain to compensate for cavity losses, we substantially improve both the quality factor and the signal intensity. Under the coupled system, Floquet modulation is induced by the alternating magnetic field, allowing for weak field detection by driving a specific hybrid mode and measuring the resulting Floquet sidebands. This miniaturized device operates at room temperature, achieving a detection limit of 121 pT/\sqrt{Hz}.

cond-mat.mes-hall↗
arXiv2026

Layer-dependent quantum transport in KV2Se2O-based altermagnetic tunnel junctions

Magnetic tunnel junction (MTJ) is the key component to enable information access and increasing number of MTJs is integrated to develop high-density spintronic devices. However, continuous miniaturization of the conventional MTJs is hindered by stray magnetic fields. Altermagnets, combining the advantages of both ferromagnets and antiferromagnets, provide a promising alternative to fabricate versatile MTJs with exotic properties, such as giant spin splitting, high intrinsic frequency, and absence of stray fields. Inspired by the altermagnetic metal candidate KV2Se2O reported recently, we design an altermagnetic tunnel junction (AMTJ) based on KV2Se2O/SrTiO3/KV2Se2O. Using density functional theory combined with non-equilibrium Green's function, we investigate the layer-dependent quantum transport properties and the tunneling magnetoresistance (TMR) of such AMTJ device. Our calculated results reveal that the transmission of the AMTJ device exhibits a pronounced oscillation behavior dependent on the number of layers of the SrTiO3 semiconductor, which is attributed to the interface configuration determined by parity of the layer number. In odd-layer devices, the electron-rich O-Se interface exhibits a smooth effective potential and enables transverse momentum (k||) transport channels, leading to enhanced transmission. In contrast, in even-layer devices, the Ti-Se interface presents a steeper effective potential, impeding quantum transport through transverse momentum (k||) channels. A giant TMR of 4.6*10^7% is predicted to be realized by using a 4-layer SrTiO3. Our findings not only provide physical understanding relevant to the quantum transport in AMTJs, but also unveil that the barrier interface engineering is a strategy to tune the magnetoelectric performance.

cond-mat.mes-hall↗
arXiv2026

LPM 1.0: Video-based Character Performance Model

Performance, the externalization of intent, emotion, and personality through visual, vocal, and temporal behavior, is what makes a character alive. Learning such performance from video is a promising alternative to traditional 3D pipelines. However, existing video models struggle to jointly achieve high expressiveness, real-time inference, and long-horizon identity stability, a tension we call the performance trilemma. Conversation is the most comprehensive performance scenario, as characters simultaneously speak, listen, react, and emote while maintaining identity over time. To address this, we present LPM 1.0 (Large Performance Model), focusing on single-person full-duplex audio-visual conversational performance. Concretely, we build a multimodal human-centric dataset through strict filtering, speaking-listening audio-video pairing, performance understanding, and identity-aware multi-reference extraction; train a 17B-parameter Diffusion Transformer (Base LPM) for highly controllable, identity-consistent performance through multimodal conditioning; and distill it into a causal streaming generator (Online LPM) for low-latency, infinite-length interaction. At inference, given a character image with identity-aware references, LPM 1.0 generates listening videos from user audio and speaking videos from synthesized audio, with text prompts for motion control, all at real-time speed with identity-stable, infinite-length generation. LPM 1.0 thus serves as a visual engine for conversational agents, live streaming characters, and game NPCs. To systematically evaluate this setting, we propose LPM-Bench, the first benchmark for interactive character performance. LPM 1.0 achieves state-of-the-art results across all evaluated dimensions while maintaining real-time inference.

cs.CV↗
arXiv2026

DecAlign: Hierarchical Cross-Modal Alignment for Decoupled Multimodal Representation Learning

Multimodal representation learning aims to capture both shared and complementary semantic information across multiple modalities. However, the intrinsic heterogeneity of diverse modalities presents substantial challenges to achieve effective cross-modal collaboration and integration. To address this, we introduce DecAlign, a novel hierarchical cross-modal alignment framework designed to decouple multimodal representations into modality-unique (heterogeneous) and modality-common (homogeneous) features. For handling heterogeneity, we employ a prototype-guided optimal transport alignment strategy leveraging gaussian mixture modeling and multi-marginal transport plans, thus mitigating distribution discrepancies while preserving modality-unique characteristics. To reinforce homogeneity, we ensure semantic consistency across modalities by aligning latent distribution matching with Maximum Mean Discrepancy regularization. Furthermore, we incorporate a multimodal transformer to enhance high-level semantic feature fusion, thereby further reducing cross-modal inconsistencies. Our extensive experiments on four widely used multimodal benchmarks demonstrate that DecAlign consistently outperforms existing state-of-the-art methods across five metrics. These results highlight the efficacy of DecAlign in enhancing superior cross-modal alignment and semantic consistency while preserving modality-unique features, marking a significant advancement in multimodal representation learning scenarios. Our project page is at https://taco-group.github.io/DecAlign.

cs.CV↗
arXiv2026

2D Ferroelectric Ruddlesden-Popper Perovskites: an Emerging Fully Electronically Controllable Shift Current and Persistent Spin Helix

Two-dimensional (2D) hybrid organic--inorganic perovskites (HOIPs) are promising candidates for next-generation optoelectronic and spintronic applications. This work systematically investigates the relationship between structural distortions and functional responses in three $C_{2v}$-symmetric Ruddlesden--Popper (RP) ferroelectric perovskites, $(4,4\text{-DFPD})_{2}\mathrm{PbI}_{4}$, $(\mathrm{DFCHA})_{2}\mathrm{PbI}_{4}$, and PEPI, using first-principles calculations combined with irreducible representation decomposition and wave-vector point-group symmetry (WPGS) analysis. The results reveal that the lead--iodide framework yields shift-current (SC) magnitudes comparable to, and in specific cases even an order of magnitude larger than, those of traditional ferroelectric oxides, with PEPI reaching a maximum of $69.16\ μ\mathrm{A}/\mathrm{V}^{2}$. The SC magnitude correlates positively with the octahedral distortion index ($D_i$), while a competition mechanism is identified between covalent bond strength and structural asymmetry, where increased average bond lengths can offset the enhancement induced by $D_i$. Regarding spintronics, $C_{2v}$ symmetry-protected persistent spin textures (PST) are identified. A transition to $C_2$-protected quasi-PST occurs in monoclinic $(4,4\text{-DFHHA})_{2}\mathrm{PbI}_{4}$, leading to a persistent spin helix (PSH) with long-distance spin transport. The synergy among ferroelectricity, SC, and PST enables nonvolatile electrical control of both photocurrent direction and spin configurations. This work provides evaluation criteria and practical guidance for designing high-performance integrated spintronic--photovoltaic devices.

cond-mat.mtrl-sci↗
arXiv2026

Modality-Aware and Anatomical Vector-Quantized Autoencoding for Multimodal Brain MRI

Learning a robust Variational Autoencoder (VAE) is a fundamental step for many deep learning applications in medical image analysis, such as MRI synthesizes. Existing brain VAEs predominantly focus on single-modality data (i.e., T1-weighted MRI), overlooking the complementary diagnostic value of other modalities like T2-weighted MRIs. Here, we propose a modality-aware and anatomically grounded 3D vector-quantized VAE (VQ-VAE) for reconstructing multi-modal brain MRIs. Called NeuroQuant, it first learns a shared latent representation across modalities using factorized multi-axis attention, which can capture relationships between distant brain regions. It then employs a dual-stream 3D encoder that explicitly separates the encoding of modality-invariant anatomical structures from modality-dependent appearance. Next, the anatomical encoding is discretized using a shared codebook and combined with modality-specific appearance features via Feature-wise Linear Modulation (FiLM) during the decoding phase. This entire approach is trained using a joint 2D/3D strategy in order to account for the slice-based acquisition of 3D MRI data. Extensive experiments on two multi-modal brain MRI datasets demonstrate that NeuroQuant achieves superior reconstruction fidelity compared to existing VAEs, enabling a scalable foundation for downstream generative modeling and cross-modal brain image analysis.

cs.CV↗
arXiv2026

No Attacker Needed: Unintentional Cross-User Contamination in Shared-State LLM Agents

LLM-based agents increasingly operate across repeated sessions, maintaining task states to ensure continuity. In many deployments, a single agent serves multiple users within a team or organization, reusing a shared knowledge layer across user identities. This shared persistence expands the failure surface: information that is locally valid for one user can silently degrade another user's outcome when the agent reapplies it without regard for scope. We refer to this failure mode as unintentional cross-user contamination (UCC). Unlike adversarial memory poisoning, UCC requires no attacker; it arises from benign interactions whose scope-bound artifacts persist and are later misapplied. We formalize UCC through a controlled evaluation protocol, introduce a taxonomy of three contamination types, and evaluate the problem in two shared-state mechanisms. Under raw shared state, benign interactions alone produce contamination rates of 57--71%. A write-time sanitization is effective when shared state is conversational, but leaves substantial residual risk when shared state includes executable artifacts, with contamination often manifesting as silent wrong answers. These results indicate that shared-state agents need artifact-level defenses beyond text-level sanitization to prevent silent cross-user failures.

cs.CL↗
arXiv2026

From molecular dynamics to kinetic models: data-driven generalized collision operators in 1D3V plasmas

We present a data-driven approach for constructing generalized collisional kinetic models for inhomogeneous plasmas in one-dimensional physical space and three-dimensional velocity space (1D-3V). The collision operator is directly learned from micro-scale molecular dynamics (MD) and accurately accounts for the unresolved particle interactions over a broad range of plasma conditions. Unlike the standard Landau operator, the present operator takes an anisotropic, non-stationary form that captures the heterogeneous collisional energy transfer arising from the many-body interactions, which is crucial for plasma kinetics beyond the weakly coupled regime. Efficient numerical evaluation is achieved through a low-rank tensor representation with $O(N \log N)$ computational complexity. The constructed kinetic equation strictly preserves conservation laws and physical constraints and therefore, enables us to develop an explicit second-order, energy-conserving scheme that ensures fully discrete conservation of mass and total energy. Numerical results demonstrate that the present model accurately predicts both transport coefficients and several 1D-3V kinetic processes compared with MD simulations across a broad range of densities and temperatures in spatially inhomogeneous settings. This work provides a systematic pathway for bridging micro-scale MD and inhomogeneous plasma kinetic descriptions where empirical models show limitation.

physics.plasm-ph↗
arXiv2026

StreamDiT: Real-Time Streaming Text-to-Video Generation

Recently, great progress has been achieved in text-to-video (T2V) generation by scaling transformer-based diffusion models to billions of parameters, which can generate high-quality videos. However, existing models typically produce only short clips offline, restricting their use cases in interactive and real-time applications. This paper addresses these challenges by proposing StreamDiT, a streaming video generation model. StreamDiT training is based on flow matching by adding a moving buffer. We design mixed training with different partitioning schemes of buffered frames to boost both content consistency and visual quality. StreamDiT modeling is based on adaLN DiT with varying time embedding and window attention. To practice the proposed method, we train a StreamDiT model with 4B parameters. In addition, we propose a multistep distillation method tailored for StreamDiT. Sampling distillation is performed in each segment of a chosen partitioning scheme. After distillation, the total number of function evaluations (NFEs) is reduced to the number of chunks in a buffer. Finally, our distilled model reaches real-time performance at 16 FPS on one GPU, which can generate video streams at 512p resolution. We evaluate our method through both quantitative metrics and human evaluation. Our model enables real-time applications, e.g. streaming generation, interactive generation, and video-to-video. We provide video results and more examples in our project website: https://cumulo-autumn.github.io/StreamDiT/

cs.CV↗
arXiv2026

Infrastructure for Valuable, Tradable, and Verifiable Agent Memory

Every API token you spend is your accumulated wealth; once you can prove its value and the effort behind it, you can resell it. As autonomous agents repeatedly call models and tools, they accumulate memories that are your intellectual property. But today these memories remain private and non-transferable, as there is no way to validate their value. We argue that agent memory can serve as an economic commodity in the agent economy, if buyers can verify that it is authentic, effort-backed, and produced in a compatible execution context. To realize this idea, we propose clawgang, which binds memory to verifiable computational provenance, and meowtrade, a market layer for listing, transferring, and governing certified memory artifacts. Together, they transform one-shot API token spending into reusable and tradable assets, enabling timely memory transfer, reducing repeated exploration, and opening a memory trade market.

cs.CR↗
arXiv2026

Retrieval-Reasoning Large Language Model-based Synthetic Clinical Trial Generation

Machine learning (ML) holds great promise for clinical applications but is often hindered by limited access to high-quality data due to privacy concerns, high costs, and long timelines associated with clinical trials. While large language models (LLMs) have demonstrated strong performance in general-purpose generation tasks, their application to synthesizing realistic clinical trials remains underexplored. In this work, we propose a novel Retrieval-Reasoning framework that leverages few-shot prompting with LLMs to generate synthetic clinical trial reports annotated with binary success/failure outcomes. Our approach integrates a retrieval module to ground the generation on relevant trial data and a reasoning module to ensure domain-consistent justifications. Experiments conducted on real clinical trials from the ClinicalTrials.gov database demonstrate that the generated synthetic trials effectively augment real datasets. Fine-tuning a BioBERT classifier on synthetic data, real data, or their combination shows that hybrid fine-tuning leads to improved performance on clinical trial outcome prediction tasks. Our results suggest that LLM-based synthetic data can serve as a powerful tool for privacy-preserving data augmentation in clinical research. The code is available at https://github.com/XuZR3x/Retrieval_Reasoning_Clinical_Trial_Generation.

cs.CL↗
arXiv2026

Agent Audit: A Security Analysis System for LLM Agent Applications

What should a developer inspect before deploying an LLM agent: the model, the tool code, the deployment configuration, or all three? In practice, many security failures in agent systems arise not from model weights alone, but from the surrounding software stack: tool functions that pass untrusted inputs to dangerous operations, exposed credentials in deployment artifacts, and over-privileged Model Context Protocol (MCP) configurations. We present Agent Audit, a security analysis system for LLM agent applications. Agent Audit analyzes Python agent code and deployment artifacts through an agent-aware pipeline that combines dataflow analysis, credential detection, structured configuration parsing, and privilege-risk checks. The system reports findings in terminal, JSON, and SARIF formats, enabling direct integration with local development workflows and CI/CD pipelines. On a benchmark of 22 samples with 42 annotated vulnerabilities, Agent Audit detects 40 vulnerabilities with 6 false positives, substantially improving recall over common SAST baselines while maintaining sub-second scan times. Agent Audit is open source and installable via pip, making security auditing accessible for agent systems. In the live demonstration, attendees scan vulnerable agent repositories and observe how Agent Audit identifies security risks in tool functions, prompts, and more. Findings are linked to source locations and configuration paths, and can be exported into VS Code and GitHub Code Scanning for interactive inspection.

cs.CR↗
arXiv2026

CNT: Safety-oriented Function Reuse across LLMs via Cross-Model Neuron Transfer

The widespread deployment of large language models (LLMs) calls for post-hoc methods that can flexibly adapt models to evolving safety requirements. Meanwhile, the rapidly expanding open-source LLM ecosystem has produced a diverse collection of models that already exhibit various safety-related functionalities. This motivates a shift from constructing safety functionality from scratch to reusing existing functionality from external models, thereby avoiding costly data collection and training procedures. In this paper, we present Cross-Model Neuron Transfer (CNT), a post-hoc method that reuses safety-oriented functionality by transferring a minimal subset of neurons from an open-source donor LLM to a target LLM. By operating at the neuron level, CNT enables modular function-level adaptation, supporting both function addition andfunction deletion. We evaluate CNT on seven popular LLMs across three representative applications: safety disalignment, alignment enhancement, and bias removal. Experimental results show that CNT achieves targeted safety-oriented functionality transfer with minimal performance degradation (less than 1% for most models), consistently outperforming five baselines, demonstrating its generality and practical effectiveness.

cs.CR↗