arXiv subjects

Chao Zhang

Publications and source records attributed to Chao Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Speech-based Psychological Crisis Assessment using LLMs

Psychological support hotlines provide critical support for individuals experiencing mental health emergencies, yet current assessments largely rely on human operators whose judgments may vary with professional experience and are constrained by limited staffing resources. This paper proposes a large language model (LLM)-based framework for automated crisis level classification, a key indicator that supports many downstream tasks and improves the overall quality of hotline services. To better capture emotional signals in spoken conversations, we introduce a paralinguistic injection method that inserts identified non-verbal emotional cues into speech transcripts, enabling LLM-based reasoning to incorporate critical acoustic nuances. In addition, we propose a reasoning-enhanced training strategy that trains the model to generate diagnostic reasoning chains as an auxiliary task, which serves as a regulariser to improve classification performance. Combined with data augmentation, our final system achieves a macro F1-score of 0.802 and an accuracy of 0.805 on the three-class classification task under 5-fold cross-validation.

cs.CL↗

arXiv2026

Talk2Agent: Benchmarking Voice Interfaces for Text Agents

Large language model (LLM) computer-use agents are typically evaluated with clean written instructions, despite speech being an increasingly popular interface for interacting with such systems. Speech input introduces an additional failure point: transcription errors can alter task-critical entities, constraints, or targets before the agent begins reasoning, while conventional ASR metrics do not directly measure whether the information required for successful execution has been preserved. We introduce Talk2Agent, a benchmark for evaluating how effectively voice interfaces convey human-spoken instructions to LLM-based computer-use agents. Talk2Agent builds human-spoken versions of tasks from WildClawBench and OSWorld and evaluates a range of voice interfaces, including dedicated ASR models, audio-capable LLMs, contextual biasing, and LLM-based ontology repair. Because repeatedly executing long-horizon computer-use tasks is costly and stochastic, we further propose an execution-free, task-conditioned evaluation framework that projects the original task grader onto prompt-addressable intentions and measures how much task-relevant information is retained after the voice interface. On WildClawBench, Talk2Agent's execution-free native projection provides a practical, execution-grounded measure of voice-interface quality, correlating with downstream task completion and improving Pearson correlation by 0.246 over WER/CER on 32 hours of real human speech.

cs.AI↗

arXiv2026

Explainable and Generalisable LLM-based Cognitive Decline Detection with Spontaneous Speech

Alzheimer's disease (AD) and mild cognitive impairment (MCI), which may precede AD, manifest early through subtle linguistic and acoustic alterations. Traditional diagnostics, however, are often resource-intensive and lack scalability for mass screening. To address these challenges, we introduce a novel bilingual speech large language model framework for automated, explainable cognitive screening. Unlike conventional pipelines that rely on error-prone automatic speech recognition, our system directly processes raw speech to learn joint acoustic-semantic representations, preserving critical prosodic cues often lost in transcription. Utilising our newly collected PUTH-AD dataset alongside multiple open-source corpora, we implemented a multi-task learning objective that simultaneously performs cognitive status classification and generates clinician-understandable natural language explanations. Our system achieved the highest average accuracy and AUROC across six dataset/task conditions, comparing three representative baselines. The system demonstrated cross-task transfer to held-out PUTH-AD task subsets, maintaining classification accuracy on an entirely unseen cognitive task without task-specific fine-tuning. Furthermore, clinician evaluation confirms that the generated explanations are both clinically relevant and largely consistent with the underlying speech evidence, supporting their potential utility in clinical interpretation. This study provides a scalable, objective, and explainable framework for speech-based cognitive screening, combining cognitive status classification with natural language explanations that clinicians can assess and verify, bridging the gap between advanced AI and clinical utility.

eess.AS↗

arXiv2026

SALMONN-duo: Adaptive Dual-System Coordination for Full-Duplex Voice Agents

Full-duplex speech large language models (LLMs) enable low-latency, natural voice interaction. However, real-world agents must also use tools and perform deliberative reasoning-operations whose variable latency and computational cost conflict with the stringent timing requirements of real-time conversation. To reconcile these demands, we propose SALMONN-duo, an adaptive dual-system voice agent inspired by dual-process theories of cognition. SALMONN-duo separates real-time interaction from deliberative computation by pairing an always-on, fast-thinking full-duplex speech LLM (system 1) with a powerful asynchronous slow-thinking LLM agent (system 2). Beyond handling real-time interaction, system 1 learns when to answer directly and when to delegate, remaining responsive during backend execution and seamlessly integrating returned information into the ongoing dialogue without exposing tool traces or losing conversational context. Evaluations on single-turn spoken question answering (QA) and multi-turn conversations demonstrate that adaptive delegation substantially improves accuracy on knowledge-intensive and multi-hop reasoning questions, while knowledge-boundary-aware training avoids unnecessary system 2 invocations. On a customized version of $τ$-Voice, SALMONN-duo further demonstrates its ability to complete environment-grounded, policy-constrained tasks through multi-turn interactions in realistic business scenarios. Finally, cost-aware reinforcement learning further enhances the trade-off between task performance and backend usage across the QA and conversation tasks, while improving task success and response safety on $τ$-Voice with an acceptable increase in the delegation rate.

cs.CL↗

arXiv2026

WeaveData: A Multimodal Data Analysis System with Self-Critiquing and Self-Evolving LLM Plans

Multimodal data analysis, which answers questions over relational tables, text, and images, has attracted growing attention in the data management community. Large language models (LLMs) enable such analysis in natural language by generating analysis plans over relational and semantic operators. However, LLM-generated plans are error-prone: a plan may silently compute something other than what was asked, fail during execution, or return a result that misses the question. This paper presents WeaveData, a multimodal data analysis system with self-critiquing and self-evolving LLM plans. First, WeaveData generates a typed logical plan for each question and critiques it step by step before execution, and it checks the executed result against the question afterwards. Second, WeaveData evolves a plan that fails or misses the question: it diagnoses the failure with the actual data, reuses the results that remain valid, and accumulates planning experience for later questions. Third, WeaveData grounds planning in a metadata knowledge graph of all modalities, clarifies ambiguous questions with the user, and backs every model judgment with evidence in an interactive notebook. We demonstrate WeaveData on two public multimodal datasets.

cs.DB↗

arXiv2026

User Association and Transmit Beamforming for Rotatable Antenna-Enabled LAE Networks

This paper investigates rotatable antenna (RA)-enabled low-altitude economy (LAE) networks, where multiple base stations (BSs) are deployed to serve both terrestrial and aerial users. Unlike conventional fixed orientation antennas, RAs can adjust their boresights to serve users at diverse azimuth and elevation angles, thereby providing three-dimensional (3D) coverage. To improve the performance of LAE networks, we formulate a network sum-rate maximization problem that jointly optimizes user association, transmit beamforming, and RA orientations. However, in such LAE networks, aerial users are often visible to multiple BSs through line-of-sight (LoS) links, making user association and interference management particularly challenging. Moreover, user association and RA orientations are closely coupled: user association affects the RA pointing directions, while the RA orientations in turn reshape the channel conditions for user association. To solve this problem, we propose an alternating optimization (AO) algorithm that integrates weighted minimum mean-square error (WMMSE) for transmit beamforming, projected gradient ascent updates for user association, and Frank-Wolfe updates for RA orientations. Furthermore, we develop a discrete candidate-scanning scheme that replaces continuous orientation optimization with a finite search to reduce computational complexity. We also propose an antenna block-based scheme to reduce the number of orientation variables. Simulation results show that the proposed joint design achieves a higher sum-rate than the fixed antenna orientation and nearest BS association benchmarks. Moreover, the candidate-scanning and block-based schemes achieve approximately 89.9% and 94.3% of the performance with high transmit power, respectively, while reducing the complexity of orientation updates.

eess.SP↗

arXiv2026

Wasserstein Convergence of ODE-Based Samplers in Decentralized Diffusion Model via Velocity Field Decomposition

Diffusion models have achieved impressive empirical success in generative tasks, and their convergence theory is now relatively well understood. Motivated by privacy and scalability, recent decentralized diffusion architectures replace a single global velocity field with multiple local experts and a routing mechanism, yielding a sampling dynamics with stochastic expert switching that falls outside standard diffusion convergence analyses. In this work, We study a decentralized diffusion framework with stochastic velocity fields and ODE-based sampling. We establish a convergence guarantee in Wasserstein-2 distance, showing that the distribution of the $N$-step discretization converges to the analytical solution at rate $\mathcal{O}(N^{-1/2}+\varepsilon)$ in $W_2$, where $\varepsilon$ captures the neural approximation errors. To our knowledge, this is the first $W_2$ convergence result for decentralized diffusion models with an ODE-based sampling scheme.

cs.LG↗

arXiv2026

Self-Evolving Time-Series Forecasting Agents with Episodic Memory and Online Policy Learning

LLM-based agents are increasingly used for time-series forecasting because they can organise contextual information, perform multi-step analysis, and guide the sequence of actions required to complete forecasting tasks. Most existing agents focus only on the current forecasting instance. However, in real-world deployments, forecasting commonly operates online, with new forecasts issued from the currently available history as the forecast origin advances and the ground-truth targets of earlier instances progressively become available. These targets provide feedback on the actions taken in earlier instances, yet existing agents generally do not preserve or utilise this information to adapt their subsequent actions. To address this limitation, we introduce FASE, a Feedback-Aware Self-Evolving forecasting agent that converts such feedback into task-specific experience for subsequent forecasting instances. FASE combines episodic memory, which retrieves relevant completed instances, with online policy learning, which summarises the feedback accumulated across instances into ranking guidance. The proposed framework is evaluated on 29 dataset configurations selected from the GIFT-Eval benchmark. Across these 29 configurations, FASE attains the strongest aggregate point forecasting performance among the evaluated methods and reduces the normalised MAE by 9.1% relative to the best individual foundation model baseline. The results further indicate that the cumulative advantage of FASE increases as delayed feedback accumulates. Together, these findings demonstrate that FASE can continually self-evolve through feedback from completed forecasting instances without updating the parameters of the LLM.

cs.LG↗

arXiv2026

Gradient estimates for the fractional $p$-Laplacian in the superquadratic regime

We establish gradient potential estimates for solutions obtained as limits of approximations (SOLA) to the fractional $p$-Laplace equation with finite signed Radon measure data. Under the assumptions $n\ge2$, $p>2$, $0 p-1$, every SOLA belongs to $W^{1,p-1}_{\mathrm{loc}}$, and its weak gradient satisfies a Wolff potential estimate at every Lebesgue point of the weak gradient. Under the additional condition $sp>n$, the solution has a continuous representative that is Fréchet differentiable at every point where the potential is finite. These results give a partial answer to questions raised by Diening and Nowak [Ann. PDE, \textbf{11}(2025)] and by Diening, Kim, Lee and Nowak~[J. Eur. Math. Soc. (JEMS), 2025] concerning gradient potential estimates for the fractional $p$-Laplacian. The proof combines homogeneous affine decay with constants independent of the affine slope and comparison estimates to obtain an affine excess recurrence. Iteration then yields the Wolff potential bound.

math.AP↗

arXiv2026

Universal Diffusion Models for Implied Volatility Surfaces: Learning Shared Dynamics Across Stocks

Modeling the dynamics of option implied volatility surface (IVS) is crucial for pricing, hedging, and risk-managing option portfolios. We develop a universal conditional diffusion model that learns to jointly generate next-day IVS increments and the underlying stock's returns. The model is trained on pooled data from 50 stocks and evaluated on a test set comprising 50 in-sample stocks and 50 out-of-sample stocks excluded from training. The training objective is primarily the minimization of MSE, with the variants that jointly impose surface smoothness, and that penalize the presence of static-arbitrage, which are all economically meaningful constraints. Across in-sample and out-of-sample stocks, our diffusion model outperforms the VolGAN benchmark in reducing arbitrage violations, improving stock risk prediction, and aligning explained variance ratios by the first three principal components. The fact that the model extrapolates well to stocks that are excluded from training indicates that the learned dynamics can be shared across stocks. These findings support a universal conditional diffusion as a credible framework for multi-stock implied volatility surface scenario generation.

q-fin.CP↗

arXiv2026

Rufus-Air: An Open LLM Post-Training Recipe

Rufus-Air is an open and reproducible post-training recipe on GLM-4.5-Air-Base (106B-A12B), organized as a serial pipeline of eight stages: SFT, Reasoning RL, Coding RL, Instruction-Following RL, General Agent, Coding Agent, Search Agent, and RLHF. We document the data, reward design, infrastructure, stage order, and stagewise results needed to reproduce the recipe. Stages progress from basic to advanced capabilities and from hard, verifiable rewards to softer judge-based signals. Training builds on open-source components and public data, much of it used as released, without new human annotation or an in-house distillation teacher. Our main findings are that (i) diverse, high-quality SFT establishes a strong capability floor; (ii) difficulty filtering keeps RL prompts within a productive learning range; (iii) reward reliability provides a practical principle for ordering stages; and (iv) infrastructure and engineering choices are part of the recipe, not just an implementation detail. Rufus-Air improves over the official GLM-4.5-Air post-trained release and is competitive with similarly sized open models.

cs.CL↗

arXiv2026

Forecast-Dojo: Replayable Environments for Benchmarking and Training LLM Forecasting Agents

We introduce Forecast-Dojo, a replayable environment for benchmarking and training LLM forecasting agents. It combines resolved prediction-market questions with dated news, allowing agents to research an event and revisit their predictions at successive historical dates. The same tasks and tools support repeated evaluation, collection of training interactions, and feedback from recorded outcomes without waiting for new events to resolve. Forecast-Dojo contains 1,568 Polymarket events, split by time into training and evaluation periods, and 18.8M dated news articles. In an evaluation of 12 models, research tools lower Brier score for all 12. Forecasts also improve as events unfold, with the largest gains at steps where more newly dated evidence is recorded. Every model still trails historical market forecasts in both Brier score and accuracy. A belief notebook carried between dates lowers research cost but does not consistently improve forecast quality. Beyond evaluation, Forecast-Dojo provides interaction trajectories and outcome feedback for agent learning, with supervised fine-tuning as a proof of concept.

cs.AI↗

arXiv2026

Coincidence double-tip scanning tunneling spectroscopy

The development of new experimental techniques for direct measurement of many-body correlations is crucial for unraveling the mysteries of strongly correlated electron systems. In this work, we propose a coincidence double-tip scanning tunneling spectroscopy (STS) that enables direct probing of spatially resolved dynamical two-body correlations of sample electrons. Unlike conventional single-tip scanning tunneling microscopy, the double-tip STS employs a double-tip scanning tunneling microscope (STM) equipped with two independently controlled tips, each biased at distinct voltages ($V_1$ and $V_2$). By simultaneously measuring the quantum tunneling currents $I_1(t)$ and $I_2(t)$ at locations $j_1$ and $j_2$, we obtain a coincidence tunneling current correlation $\overline{\langle I_1(t) I_2(t)\rangle}$. Differentiating this coincidence tunneling current correlation with respect to the two bias voltages yields a coincidence dynamical conductance. Through the development of a nonequilibrium theory, we demonstrate that this coincidence dynamical conductance is proportional to a contour-ordered second-order current correlation function. For the sample electrons in a nearly free Fermi liquid state, the coincidence dynamical conductance captures two correlated dynamical electron propagation processes: (i) from $j_1$ to $j_2$ (or vice versa) driven by $V_1$, and (ii) from $j_2$ to $j_1$ (or vice versa) driven by $V_2$. For the sample electrons in a superconducting state, additional propagation channels emerge from the superconducting condensate, coexisting with the above normal electron propagation processes. Thus, the coincidence double-tip STS provides direct access to spatially resolved dynamical two-body correlations, offering a powerful tool for investigating strongly correlated electron systems.

cond-mat.str-el↗

arXiv2026

Decoupling Perception from Description: Computation-Grounded Representation Alignment between Multivariate Time Series and Language

Training multimodal models to align time series with language runs into a self-supervision trap. The usual recipe asks an LLM to read a series and write a description, so label quality is capped by the perceptual skill the model is supposed to learn. The data can never teach more than the labeler already knows. A second gap makes this worse: most datasets use a single variable, but the patterns that matter (cross-channel correlation, lead-lag structure, co-occurring anomalies) appear only with several variables, right where the labeling LLM's limits are most exposed. These two problems create a trilemma: existing methods are reliable, realistic, or scalable, but none achieves all three. We resolve this by decoupling perception from description. Deterministic code computes a set of statistics from real, open-source multivariate series; the LLM verbalizes those precomputed facts. Perception, which LLMs do poorly, is handled by computation, while the LLM handles expression. This produces CGTime, our 4B-parameter computation-grounded time-series-language model. CGTime outperforms far larger general-purpose models on multivariate understanding tasks: it attains the best multivariate fact score on our held-out benchmark (0.283 vs. 0.173 for GPT-4o-mini and 0.203 for GPT-5.4-nano), a gap that survives Holm-corrected paired significance tests against every baseline. It also states verifiable numerical facts in generated captions more accurately and covers a broader range of statistical properties.

cs.LG↗

arXiv2026

H-VLA: Hierarchical Vision-Language-Action Model with Key-Action Reasoning and Motion Planning in a Unified Action Space

Vision-Language-Action (VLA) models have shown strong potential for robotic manipulation, but many existing methods still rely on direct mappings from language and visual observations to dense actions. This formulation can weaken the semantic reasoning capability inherited from pre-trained Vision-Language Models (VLMs), which are mainly optimized for visual-linguistic understanding rather than low-level control, and becomes fragile under spatial variations, including changes in object positions, scene layouts, robot embodiments, and camera viewpoints. To address these limitations, we propose H-VLA, a hierarchical VLA framework that decouples high-level key-action reasoning from low-level motion generation. H-VLA combines a Key-Action Model for predicting a key-action as the next manipulation subgoal, a Motion Planning Model for generating dense future actions conditioned on the predicted key-action, and a Unified Camera-Centric Action Space for consistent representation across datasets, embodiments, and viewpoints. We further adopt a two-stage training strategy that emphasizes key-action reasoning during pre-training and dense motion generation during fine-tuning. Experiments show that H-VLA achieves strong performance on SimplerEnv, reaching 91% on Google Robot visual matching, 84% on Google Robot variant aggregation, and 81% on WidowX visual matching. On Agilex real-robot tasks, H-VLA improves over the strongest baseline by 10, 47, and 16 percentage points under in-distribution, out-of-distribution position, and out-of-distribution scene/object settings, respectively.

cs.RO↗

arXiv2026

Mem-World: Memory-Augmented Action-Conditioned World Models for Persistent Robot Manipulation

Action-conditioned world models have emerged as a promising paradigm for robot learning, offering a scalable alternative to costly real-world experimentation by generating action-consistent video rollouts. However, persistent world modeling remains challenging in manipulation: frequent end-effector occlusions and rapid wrist-camera motion make the current observation insufficient for predicting future views, causing models to forget or hallucinate scene details seen in earlier frames. Existing memory retrieval strategies often fail to identify informative history in dynamic manipulation scenarios. To address this limitation, we propose Mem-World, a memory-augmented multi-view action-conditioned world model. At its core, we present W-VMem, a 4D wrist-view-centered surfel-indexed memory that anchors historical observations to temporally evolving surface elements. By explicitly modeling when and where scene elements are observed, W-VMem enables geometry-aware retrieval of relevant history frames conditioned on future actions. During generation, relevant history frames are selected via surfel-based rendering and scoring, providing informative and non-redundant context for prediction. Extensive experiments show that Mem-World generates persistent rollouts in complex manipulation scenarios, enables more reliable policy evaluation than Ctrl-World, improving the Pearson correlation with real-world performance by 14.5\%, and supports effective policy improvement through synthetic data generation, increasing success rates from 58\% to 72\% on long-horizon tasks.

cs.CV↗

arXiv2026

CoFi-CLM: A Coarse-to-Fine Channel Language Model for Finite-Bit CSI Feedback

In frequency-division duplex massive multiple-input multiple-output (MIMO) systems, the user equipment (UE) must convey high-dimensional downlink channel state information (CSI) under a stringent finite-bit feedback budget. Deep learning (DL) has emerged as a powerful tool for CSI compression due to its ability to capture complex channel correlations and learn compact CSI representations. However, recovering fine-scale channel structure from limited feedback remains challenging. To address this challenge, we propose the Coarse-to-Fine Channel Language Model (CoFi-CLM), a large AI model deployed at the base station (BS) that explicitly learns dependencies between coarse and fine CSI representations. Specifically, the UE reports coarse-token indices from a learned codebook, while CoFi-CLM predicts distributions over unreported fine-token indices conditioned on this feedback in a single Transformer forward pass. A dual-path decoder fuses the direct coarse reconstruction with the generated fine-scale reconstruction. Concentrating fine-token prediction and fusion at the BS allows the CLM capacity to scale without increasing the computational or storage cost at the UE. Across 64 to 160 feedback bits, CoFi-CLM consistently outperforms the compared methods on both seen and unseen scenarios. At 128 bits, it improves the normalized mean square error (NMSE) over the recent large AI model baseline by approximately 1 dB on both sets. The performance gap between seen and unseen scenarios further supports its generalization under the considered channel model. The source code is publicly available at https://github.com/doovvv/CoFi-CLM.

eess.SP↗

arXiv2026

Polarization-Aware Rotatable Antennas for RIS-Empowered Symbiotic Radios

This paper investigates a dual-polarized reconfigurable intelligent surface (DP-RIS)-empowered symbiotic radio (SR) system with rotatable antennas (RAs). By reconfiguring antenna orientations, RAs can steer their radiation patterns toward desired directions, thereby effectively mitigating the double-fading effect in RIS-assisted cascaded links. However, antenna rotation not only changes the radiation direction but also alters the local polarization bases, which may result in polarization mismatch and degrade the achievable gain. This motivates a joint spatial-polarization design that simultaneously exploits directional radiation gain and polarization matching. Specifically, we formulate a transmit power minimization problem that jointly optimizes digital beamforming, RA rotations, transceiver polarization states, and DP-RIS phase shifts, subject to the primary and secondary rate requirements as well as interference temperature constraints for non-SR users. To solve this non-convex problem, we develop an alternating optimization algorithm that integrates semidefinite programming, difference-of-convex programming, and Riemannian conjugate gradient methods. Moreover, to reduce hardware and computational complexity for practical deployment, we propose two low-complexity RA designs, namely a subarray-wise shared rotation design and a discrete rotation codebook design. Simulation results show that the proposed polarization-aware RA design significantly reduces the required transmit power compared with fixed-orientation and polarization-unaware benchmark schemes. Moreover, both the proposed low-complexity RA designs achieve comparable performance with reduced rotation complexity.

eess.SP↗