arXiv subjects

Yichen Zhang

Publications and source records attributed to Yichen Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Workflow-GYM: Towards Long-Horizon Evaluation of Computer-use Agentic tasks in Real-World Professional Fields

Recent years have witnessed the rapid evolution of AI agents toward handling increasingly complex, real-world tasks. However, existing benchmarks rarely evaluate whether agents can operate graphical user interfaces to complete long-horizon, high-value professional workflows across diverse domains. Current GUI benchmarks still predominantly focus on general-purpose software, relatively simple applications, and short-horizon tasks, leaving it largely unknown whether modern agents can follow user instructions to autonomously operate domain-specific professional software and accomplish economically valuable work in an end-to-end manner. To bridge this gap, we introduce Workflow-GYM, a benchmark for long-horizon GUI tasks centered on professional domains and specialized software environments. Through extensive experiments on state-of-the-art models, we find that even the strongest models achieve only slightly above 30% success rates, highlighting that professional long-horizon GUI workflows remain highly challenging for current GUI agents. Further analysis reveals that current agents struggle to maintain long-horizon workflow consistency, frequently exhibiting workflow stage omission, error propagation, objective drift, and insufficient understanding of professional software environments. Our findings provide important insights into the limitations of current agent systems and suggest key directions for the next generation of GUI-agent research.

cs.AI↗

arXiv2026

StatEval: A Comprehensive Benchmark for Large Language Models in Statistics

Despite rapid advances in large language models (LLMs), statistical reasoning remains underrepresented in existing LLM benchmarks, which often do not reflect the layered, proof-driven nature of real statistical practice. To address this gap, we introduce \textbf{StatEval}, the first large-scale benchmark for statistical reasoning across curricular and research-level settings. StatEval includes over 100,000 curated problems, with 20,000+ foundational questions spanning undergraduate and graduate curricula and 80,000+ research-level proof tasks extracted from leading statistical journals. To construct StatEval, we develop \textbf{TRACE} (Topology and Reasoning-Aware Context Extractor), a multi-agent pipeline with human-in-the-loop validation that converts unstructured academic texts into self-contained theorem-level reasoning tasks. We also propose an Adaptive Process-Based Scoring Pipeline for complex statistical proofs, enabling fine-grained evaluation beyond final-answer matching. Experiments show that while LLMs perform reasonably on foundational tasks, they struggle with rigorous research-level reasoning. Beyond evaluation, StatEval serves as a resource for improving reasoning, as retrieval-augmented generation and domain-specific alignment consistently enhance performance. Together, these results establish StatEval as both a benchmark and an infrastructure for advancing statistical reasoning in LLMs.

cs.CL↗

arXiv2026

Large Language Model Agents Are Not Always Faithful Self-Evolvers

Self-evolving large language model (LLM) agents continually improve by accumulating and reusing past experience, yet it remains unclear whether they faithfully rely on that experience to guide their behavior. We present the first systematic investigation of experience faithfulness, the causal dependence of an agent's decisions on the experience it is given, in self-evolving LLM agents. Using controlled causal interventions on both raw and condensed forms of experience, we comprehensively evaluate four representative frameworks across 13 LLM backbones and 9 environments. Our analysis uncovers a striking asymmetry: while agents consistently depend on raw experience, they often disregard or misinterpret condensed experience, even when it is the only experience provided. This gap persists across single- and multi-agent configurations and across backbone scales. We trace its underlying causes to three factors: the semantic limitations of condensed content, internal processing biases that suppress experience, and task regimes where pretrained priors already suffice. These findings challenge prevailing assumptions about self-evolving methods and underscore the need for more faithful and reliable approaches to experience integration.

cs.CL↗

arXiv2026

HOTDISK. Finding Massive Protostellar Disks with Water and Refractory Molecular Species

We present high-angular-resolution ($\sim0.05^{\prime\prime}$, $\sim 60-250$ au) ALMA Band~6 observations from the HOTDISK project (Hot-Origin Tracer survey of DISKs of massive protostars) aimed at investigating the "hot-disk" chemical pattern traced by vibrationally excited water, NaCl, SiS, and SiO in the innermost regions around massive protostars. Ten targets were selected based on strong CH$_3$CN emission exhibiting clear rotational signatures and centrally concentrated SiO emission from lower-resolution observations. We detect vibrationally excited water emission toward 7 of the 10 sources. In all detections, the blueshifted and redshifted components are compact and located on opposite sides of the 1.3 mm continuum peak, with velocity gradients approximately perpendicular to the outflow axes, consistent with rotation on disk scales. Emission from NaCl and SiS is detected toward 5 of these 7 sources and exhibits similar kinematics, further supporting the presence of compact rotating structures. In contrast, commonly used hot-core tracers (e.g., CH$_3$CN and SO$_2$) primarily probe larger-scale envelope gas. These results demonstrate that vibrationally excited water, NaCl, and SiS are powerful tracers of disk structures on $\sim$100 au scales, when observed at sufficient angular resolution and sensitivity. The high detection rate suggests that hot-disk chemical patterns -- and thus compact rotating disks -- are common in massive star-forming regions, at least among sources with well-developed rotating envelopes.

astro-ph.GA↗

arXiv2026

Binomial Smoothing for Inventory and Information Control in Supply Chains

In many decentralized supply chains, upstream firms do not observe market demand directly and instead infer downstream conditions from the order stream. A retailer's replenishment policy therefore plays a dual role: it governs inventory replenishment and shapes the information available for upstream forecasting. This creates a fundamental trade-off. Smoother orders improve upstream predictability, but delaying the response to demand can increase downstream inventory costs. We study how a retailer should optimally smooth demand in a two-tier supply chain with one retailer and one manufacturer when the manufacturer forecasts future orders from the retailer's order history. We propose Binomial Smoothing, a class of replenishment policies that implements delayed demand response by spreading each unit of demand over a finite horizon using binomial weights. The class is interpretable, easy to calibrate, and analytically tractable. Under weakly stationary Gaussian demand satisfying mild regularity conditions, we show that, for any fixed smoothing horizon, the Binomial policy minimizes the manufacturer's forecast error among all policies with the same degree of smoothing. It remains invertible, so the manufacturer can recover demand history from observed orders. More generally, Binomial Smoothing achieves a constant-factor approximation guarantee relative to an optimal policy. Our results yield a broader insight: replenishment policies should be designed not merely to reduce order variance, as in the traditional bullwhip measure, but to reduce the unpredictable component of orders. Carefully designed smoothing can improve supply-chain performance and partially substitute for information sharing, providing a concrete mechanism for coordination without collaboration.

stat.AP↗

arXiv2026

The 1st PortraitCraft Challenge: A CVPR 2026 Workshop Competition on Portrait Composition Understanding and Generation

This paper presents an overview of the inaugural PortraitCraft Challenge, held as one of the official competitions at CVPR 2026. The challenge focuses on portrait composition understanding and generation, aiming to advance AI research in portrait aesthetics analysis and controllable image synthesis. Unlike existing datasets and tasks that primarily focus on global aesthetic scoring, PortraitCraft introduces a unified evaluation framework comprising two complementary tracks. Track 1 requires models to perform structured portrait composition understanding, and Track 2 requires models to generate portrait images from structured composition descriptions under explicit compositional constraints. To support the challenge, we constructed and publicly released a large-scale portrait composition dataset consisting of approximately 50,000 curated real portrait images, providing multi-level supervision. This report describes the challenge setup, evaluation protocols, dataset composition, and final results, along with an analysis of the technical characteristics of the submitted solutions. The PortraitCraft Challenge provides a standardized and reproducible platform for research on portrait composition understanding and generation, and is expected to foster further progress in the fields of portrait aesthetics and controllable image generation.

cs.CV↗

arXiv2026

Generalized Rank Regression

Rank regression offers robustness to outliers and heavy-tailed response distributions, invariance to monotonic transformations, and improved efficiency under non-Gaussian errors, making it a versatile tool for analyzing complex data. This paper introduces Generalized Rank Regression (GRR), an extension of classical rank-based methods that accommodates non-monotonic score functions. While aimed at enhancing the statistical efficiency of robust estimators, this generalization results in a potentially non-convex and non-smooth objective function, presenting challenges for both theoretical analysis and algorithmic implementation. We derive a non-asymptotic Bahadur representation of the proposed estimator and establish its asymptotic normality under mild conditions. To address the optimization challenges, we propose a new two-stage sub-gradient descent algorithm that enables efficient computation of GRR estimators with desirable statistical properties. Furthermore, we develop a multiplier bootstrap procedure for conducting statistical inference. A close connection between GRR and variants of quantile regression is uncovered, which demonstrates that GRR and composite quantile regression share asymptotically equivalent variances. The advantages of GRR are illustrated through extensive simulation studies and a real data application.

stat.ME↗

arXiv2026

MuMuTestUp: Mutation-based Multi-Agent Test Case Update

Modern software systems evolve rapidly under CI/CD practices, where tests are critical for quality. However, substantial code changes often render existing test cases obsolete, causing pipeline disruptions, reduced productivity, and compromised quality. Recent automatic test update approaches leverage LLMs to refine test cases via execution feedback and exact-matching context retrieval, prioritizing executability and line coverage but suffering three limitations: (1) neglecting test assertion adequacy, weakening fault detection; (2) relying on coarse line coverage instead of specific uncovered lines/branches; (3) using exact-matching retrieval, which fails for LLM hallucinated queries. To address these, we propose MuMuTestUp, a mutation-guided multi-agent framework with three specialized agents: Mutation Analysis (strengthens assertions via surviving mutants), Coverage Analysis (generates targeted repair instructions for uncovered lines/branches), and Semantic Retrieval (handles hallucinations via semantic-similarity search). We also construct PRBENCH, a 571-sample pull-request-level dataset from 10 open-source Java projects (validated for cross-commit update scenarios). Evaluations against state-of-the-art baselines use both open-source (Deepseek-V3.2) and closed-source (GPT-4.1) LLMs.

cs.SE↗

arXiv2026

U-STS-LLM A Unified Spatio-Temporal Steered Large Language Model for Traffic Prediction and Imputation

The efficient operation of modern cellular networks hinges on the accurate analysis of spatio-temporal traffic data. Mastering these patterns is essential for core network functions, chiefly forecasting future load to pre-empt congestion and imputing missing values caused by sensor failures or transmission errors to ensure data continuity. While deeply connected, forecasting and imputation have historically evolved as separate sub-fields. The dominant paradigm, Spatio-Temporal Graph Neural Networks (STGNNs), while effective, are often specialized, computationally intensive, and exhibit limited generalization. Concurrently, adapting large pre-trained language models (LLMs) offers a powerful alternative for sequence modeling, yet existing approaches provide weak structural guidance, leading to unstable convergence and a narrow focus on forecasting. To bridge these gaps, we propose U-STS-LLM, a unified framework built on a spatio-temporally steered LLM. Our core innovation is a Dynamic Spatio-Temporal Attention Bias Generator that synthesizes a persistent functional graph with transient nodal states to explicitly steer the LLM's attention. Coupled with a partially frozen backbone tuned via Low-Rank Adaptation (LoRA) and a Gated Adaptive Fusion mechanism, the model achieves stable, parameter-efficient adaptation. Trained under a unified multi-task objective, U-STS-LLM learns a holistic data representation. Extensive experiments on real-world cellular datasets demonstrate that U-STS-LLM establishes new state-of-the-art performance in both long-horizon forecasting and high-missing-rate imputation, while maintaining remarkable training efficiency and stability, offering a novel blueprint for harnessing foundation models in structured, non-linguistic domains.

cs.LG↗

arXiv2026

Rethinking Experience Utilization in Self-Evolving Language Model Agents

Self-evolving agents improve by accumulating and reusing experience from past interactions. Existing work has largely focused on how experience is constructed, represented, and updated, while paying less attention to how experience should be used during runtime decision-making. As a result, most agents rely on rigid usage strategies, either injecting experience once at initialization or at every step, without considering whether it is needed for the current decision. This paper studies experience utilization as a critical design dimension of self-evolving agents. We ask whether agents benefit from interweaving experience use with decision-making, so that experience is invoked only when additional guidance is needed. To examine this question, we introduce {ExpWeaver}, a lightweight instantiation that leaves experience construction unchanged and modifies only runtime utilization by exposing experience as an optional resource during reasoning. Across four representative frameworks, seven LLM backbones, and three types of environments, ExpWeaver consistently achieves the best performance among different utilization strategies. Reinforcement learning experiments further show that this behavior can be amplified through training. Usage-pattern, causal ablation, and entropy-based analyses reveal that ExpWeaver enables agents to invoke experience selectively, at beneficial decision points, and under higher reasoning uncertainty. Overall, our findings call for a shift from merely studying \emph{what} experience to store toward understanding \emph{how} and \emph{when} experience should enter decision-making.

cs.CL↗

arXiv2026

The Keplerian disk, envelope, and streamers surrounding an early O-type protostar in the Sagittarius C cloud of the Central Molecular Zone

Disk-mediated accretion is central to theories of massive star formation, setting the initial conditions for their evolution. Yet observations of Keplerian disks around early O-type protostars remain scarce, as they are often blended into complex surrounding structures. We report ALMA Band 6 observations (300 au resolution) of an accretion disk surrounding a high-mass protostar in the Sagittarius C (Sgr C) cloud in the Central Molecular Zone (CMZ) around the Galactic Center. We identify spectral lines and analyze the spatial distribution of the emission of the complex organic molecules. We use a dynamical model with an inner Keplerian disk and an outer free-fall envelope to fit the three-dimensional position-position-velocity data of the stacked CH$_3$OCHO molecular lines and constrain the mass of the central protostar to be $\sim40^{+2}_{-3} M_{\odot}$. The fitting results additionally show that the disk has a centrifugal radius at about 1300 au. Considering the infall velocity, radius, and mass of the envelope, we estimate the accretion rate from the envelope onto the disk to be $\sim7\times 10^{-3}\ M_{\odot}\,\mathrm{yr^{-1}}$. We also identify spiral-like structures in the disk that can be described by free-falling streamers. Our results highlight the critical role of accretion disks and streamers in the mass accumulation of early O-type stars in the CMZ.

astro-ph.GA↗

arXiv2026

PI-TTA: Physics-Informed Source-Free Test-Time Adaptation for Robust Human Activity Recognition on Mobile Devices

Source-free test-time adaptation (TTA) is appealing for mobile and wearable sensing because it enables on-device personalization from unlabeled test streams without centralizing private data. However, sensor-based human activity recognition (HAR) poses challenges that are less pronounced in standard vision benchmarks: behavioral inertial streams are temporally correlated and often exhibit within-session shifts caused by sensor rotation, placement change, and sampling-rate drift. Under this streaming non-i.i.d. setting, widely used vision-style TTA objectives can become unstable, leading to overconfident errors, representation collapse, and catastrophic forgetting. We propose PI-TTA, a lightweight source-free adaptation framework that stabilizes online updates through three physics-consistent constraints: gravity consistency, short-horizon temporal continuity, and spectral stability. PI-TTA updates the same small parameter subset as strong source-free baselines and incurs only modest overhead, making it suitable for on-device deployment. Experiments on USCHAD, PAMAP2, and mHealth under long-sequence stress tests and factorized shift protocols show that PI-TTA mitigates the severe degradation observed in confidence-driven baselines and preserves stable adaptation under sustained streaming conditions. It improves long-sequence accuracy by up to 9.13% and reduces physical-violation rates by 27.5%, 24.1%, and 45.4% on USCHAD, PAMAP2, and mHealth, respectively. These results demonstrate that physics-informed adaptation can improve accuracy, stability, and deployment reliability for real-world mobile sensing systems.

cs.AI↗

arXiv2026

On Safety Risks in Experience-Driven Self-Evolving Agents

Experience-driven self-evolution has emerged as a promising paradigm for improving the autonomy of large language model agents, yet its reliance on self-curated experience introduces underexplored safety risks. In this study, we investigate how experience accumulation and utilization in self-evolving agents affect safety performance across web-based and embodied environments. Notably, experience gathered solely from benign tasks can still compromise safety in high-risk scenarios. Further analysis attributes this degradation to the execution-oriented nature of accumulated experience, which reinforces agents' tendency to act rather than refuse. In more realistic settings where agents encounter both benign and harmful tasks, refusal-related experience mitigates safety decline but induces over-refusal, revealing a fundamental safety-utility trade-off. Overall, our findings expose inherent limitations of current self-evolving agents and call for more principled strategies to ensure safe and reliable adaptation.

cs.CL↗

arXiv2026

High-resolution long-range 3D single-photon imaging with a compact SPAD array

High-resolution three-dimensional imaging under photon-starved conditions remains challenging. Here, we demonstrate a high-resolution long-range 3D single-photon imaging system based on a digital micromirror device (DMD) and a compact 64 multiply 64 single-photon avalanche diode (SPAD) array. By combining high-resolution spatial modulation with parallel time-resolved detection, the system extends the effective spatial sampling beyond the native detector format while preserving depth information through time-of-flight measurement. In outdoor experiments at a stand-off distance of 670 m, we achieved 3D reconstruction of natural targets with an effective spatial resolution of 256 multiply 256. These results validate the proposed method as an effective approach for high-resolution long-range 3D single-photon imaging using compact SPAD arrays.

physics.optics↗

arXiv2026

Policy-Aware Design of Large-Scale Factorial Experiments

Digital firms routinely run many online experiments on shared user populations. When product decisions are compositional, such as combinations of interface elements, flows, messages, or incentives, the number of feasible interventions grows combinatorially, while available traffic remains limited. Overlapping experiments can therefore generate interaction effects that are poorly handled by decentralized A/B testing. We study how to design large-scale factorial experiments when the objective is not to estimate every treatment effect, but to identify a high-performing policy under a fixed experimentation budget. We propose a two-stage design that centralizes overlapping experiments into a single factorial problem and models expected outcomes as a low-rank tensor. In the first stage, the platform samples a subset of intervention combinations, uses tensor completion to infer performance on untested combinations, and eliminates weak factor levels using estimated marginal contributions. In the second stage, it applies sequential halving to the surviving combinations to select a final policy. We establish gap-independent simple-regret bounds and gap-dependent identification guarantees showing that the relevant complexity scales with the degrees of freedom of the low-rank tensor and the separation structure across factor levels, rather than the full factorial size. In an offline evaluation based on a product-bundling problem constructed from 100 million Taobao interactions, the proposed method substantially outperforms one-shot tensor completion and unstructured best-arm benchmarks, especially in low-budget and high-noise settings. These results show how centralized, policy-aware experimentation can make combinatorial product design operationally feasible at platform scale.

stat.ML↗

arXiv2026

DreamerAD: Efficient Reinforcement Learning via Latent World Model for Autonomous Driving

We introduce DreamerAD, the first latent world model framework that enables efficient reinforcement learning for autonomous driving by compressing diffusion sampling from 100 steps to 1 - achieving 80x speedup while maintaining visual interpretability. Training RL policies on real-world driving data incurs prohibitive costs and safety risks. While existing pixel-level diffusion world models enable safe imagination-based training, they suffer from multi-step diffusion inference latency (2s/frame) that prevents high-frequency RL interaction. Our approach leverages denoised latent features from video generation models through three key mechanisms: (1) shortcut forcing that reduces sampling complexity via recursive multi-resolution step compression, (2) an autoregressive dense reward model operating directly on latent representations for fine-grained credit assignment, and (3) Gaussian vocabulary sampling for GRPO that constrains exploration to physically plausible trajectories. DreamerAD achieves 87.7 EPDMS on NavSim v2, establishing state-of-the-art performance and demonstrating that latent-space RL is effective for autonomous driving.

cs.LG↗

arXiv2026

The SOMA Atomic Outflow Survey. I. An Atomic OI and Highly Ionized OIII Outflow from Massive Protostar G11.94-00.62

Massive stars regulate galaxy evolution and star formation through their physical and chemical feedback, but their formation remains poorly understood. Accretion-powered outflows provide important diagnostics of massive star formation. We present first results from the SOMA Atomic Outflow Survey, a far-infrared massive star formation survey using the FIFI-LS instrument on SOFIA. We report detection of \OIII\ $^3P_2\rightarrow^3P_1$ emission at 52 \micron\ from the massive protostar G11.94-0.62, tracing highly ionized gas. We also detect \OI\ $^3P_2\rightarrow^3P_1$ and $^3P_1\rightarrow^3P_0$ at 63 and 145 \micron\ tracing atomic gas, as well as CO $J=14\rightarrow13$ at 186 \micron\ from highly excited molecular gas. The \OIII\ and \OI\ lines exhibit large line widths ($\sim200$ and $\sim40-80$ \kms, respectively) and their morphologies are consistent with a wide-angle bipolar outflow. The properties of molecular tracers ($^{12}$CO, $^{13}$CO, C$^{18}$O, H$_2$CO, and CH$_3$OH) observed with ALMA support this interpretation. Ionized nebula and PDR modeling imply an ionized outflow mass flux of $\sim8\times10^{-5}\:M_\odot$ yr$^{-1}$ and an atomic outflow mass flux of $\sim5\times10^{-6}\:M_\odot$ yr$^{-1}$, while the molecular outflow traced by CO has an implied mass flux of $\sim3\times10^{-4}\:M_\odot$ yr$^{-1}$. The mass and momentum flux in the ionized outflow are consistent with the primary disk wind, while the molecular component is mainly swept-up, secondary outflow gas. We also observe G11.94-0.62 with the LBT in the near-infrared, potentially tracing the base of wide-angle outflow cavities. SED modeling implies a protostellar mass $m_* = 22.4^{+21}_{-11}\:M_\odot$, while the \OIII\ emission implies $m_*\gtrsim30\:M_\odot$ and that the protostar is in the final stages of its accretion.

astro-ph.SR↗

arXiv2026

Latent-WAM: Latent World Action Modeling for End-to-End Autonomous Driving

We introduce Latent-WAM, an efficient end-to-end autonomous driving framework that achieves strong trajectory planning through spatially-aware and dynamics-informed latent world representations. Existing world-model-based planners suffer from inadequately compressed representations, limited spatial understanding, and underutilized temporal dynamics, resulting in sub-optimal planning under constrained data and compute budgets. Latent-WAM addresses these limitations with two core modules: a Spatial-Aware Compressive World Encoder (SCWE) that distills geometric knowledge from a foundation model and compresses multi-view images into compact scene tokens via learnable queries, and a Dynamic Latent World Model (DLWM) that employs a causal Transformer to autoregressively predict future world status conditioned on historical visual and motion representations. Extensive experiments on NAVSIM v2 and HUGSIM demonstrate new state-of-the-art results: 89.3 EPDMS on NAVSIM v2 and 28.9 HD-Score on HUGSIM, surpassing the best prior perception-free method by 3.2 EPDMS with significantly less training data and a compact 104M-parameter model.

cs.CV↗