arXiv subjects

Yuheng Wang

Publications and source records attributed to Yuheng Wang.

At least 19 recordsLinked to original sources

arXiv2026

Agentic University: Collaborative Agents for Personalized Learning

Large language models support personalized tutoring, content generation, and adaptive learning paths. However, capability development requires these resources and interactions to form a coherent learning process that evolves with the learner. A central challenge is to align goals, activities, and assessment while using learning evidence to decide what should follow. We introduce Human-Participatory Multi-Agent Organization and instantiate it in Agentic University (AU). In AU, learners participate as organization members and work with specialized agents to maintain a revisable Capability Program. The program connects personal goals, learning activities across modalities, and layered assessment, while learner feedback and performance guide subsequent support. Our evaluation shows advantages in program organization, assessment design, and revision quality. Users achieved larger mean assessment gains (12.19 versus 5.61 points on a 50-point scale) and reported stronger cognitive support. This work contributes an organizational approach to adaptive learning that makes learner participation a continuing basis for multi-agent coordination. The Agentic University project is available at \href{https://github.com/AgenticOrgLab/AgenticUniversity.git}.

cs.AI↗

arXiv2026

Quantifying the Impact of Motion on 2D Gaze Estimation in Real-World Mobile Interactions

Mobile gaze tracking on handheld devices commonly relies on \emph{appearance-based 2D gaze estimation} from the device's \emph{front-facing RGB camera}. Although this approach makes eye tracking more accessible by avoiding dedicated infrared hardware, maintaining accuracy remains difficult because the spatial relationship among the user's face, the camera, and the screen changes continuously during handheld use. This paper examines how motion-related geometry change affects a representative front-camera RGB 2D gaze-estimation pipeline. We report two complementary within-subject user studies ($n=10$) that capture handheld behaviour, calibration data, and gaze-estimation error during everyday-style mobile tasks across conditions ranging from lying and sitting to walking and maze navigation. Rather than treating motion as a single binary factor, we analyse within-task versus task-switch behaviour, calibration-test mismatch, and condition-specific motion signatures. Behavioural results show that head-to-camera distance is comparatively stable during sustained within-task use, while task switches introduce much larger geometry fluctuations, with head-distance standard deviations of approximately 7.42--10.10\,cm across conditions. In the gaze-evaluation study, one-off calibration reduced average RMSE from 4.49\,cm to 3.44\,cm; under this regime, the global effect of motion condition on RMSE was not statistically reliable, whereas clearer patterns appeared in task-specific and motion-specific calibration analyses, where better calibration-test match reduced error.

cs.HC↗

arXiv2026

Collaborative Learning and Reasoning over Model Slices

Large language models (LLMs) excel across diverse tasks, benefiting from knowledge priors, in-context reasoning, and test-time computation. Despite these strengths, single-model reasoning remains vulnerable to error propagation, as later steps depend on the model's earlier outputs, leaving reasoning blind spots difficult to independently reassess. To address this limitation, we develop \textit{collaborative learning}, encompassing the construction of heterogeneous collaborators and the learning of effective cross-model collaboration. Specifically, \textit{model slicing} replaces a large model with smaller, heterogeneous models under a comparable parameter budget, enabling prior reasoning to be reassessed across distinct model parameterizations. \textit{Pyramid supervision} combines protocol distillation with step-level supervision to refine reasoning and interaction decisions throughout collaboration. We evaluate our approach across four task categories: multi-hop question answering, mathematical reasoning, code generation, and open-ended constrained generation. The trained heterogeneous models achieve comparable or better task performance than the original large model and established multi-agent baselines, including reinforcement-learning-based methods, while also achieving higher inference efficiency than the multi-agent baselines and exhibiting differentiated collaboration patterns across tasks. These findings show that learned collaboration turns model heterogeneity into complementary reasoning capabilities.

cs.AI↗

arXiv2026

OctoNest: Adaptive Cross-Device Execution through Stateful Control

Computer use agents are expanding from single-device operation toward cross-device systems that coordinate tasks across heterogeneous environments. Execution conditions are often only partially known at planning time and revealed through interaction. Failures may require intra-device modality switching or inter-device reassignment; failing to distinguish these cases can lead to repeated failures or premature termination. However, existing systems primarily scale up single-device agents without sufficiently distinguishing device-level and modality-specific execution conditions. We propose OctoNest, which coordinates stateful cross-device orchestration and iterative device-local modality control. Device Agents refine subtasks and select modalities, while an Orchestrator uses execution feedback to revise plans and device assignments. We also introduce CAPEBench, comprising 158 instances from 23 cross-device seed tasks with controlled perturbations. OctoNest leads all three quality metrics, improving Perfect Pass over the strongest baseline by 18.35 percentage points and reducing token cost per perfect pass by 39.8\%. Further analyses support the complementary roles of local refinement and global revision and demonstrate CAPEBench's ability to distinguish control limitations under changing execution conditions.

cs.CL↗

arXiv2026

CoEvoke: Evoking Multi-Agent Collaboration for Trajectory Collection

Training multi-agent collaboration requires trajectories of checking, feedback, and repair, yet task-solving rollouts may end in immediate approval or repeat unsuccessful attempts without useful correction. We introduce CoEvoke, a cross-domain framework that evokes these interactions during trajectory collection. Controlled perturbations create recoverable difficulties, while restricted reference-based guidance directs investigation. The resulting dataset contains 4,567 reconstructable trajectories over 4,130 software-engineering and reasoning tasks, linking public interactions to evidence, decisions, and task-native outcomes. Separate experiments over 2,287 same-task pairs show feedback increases of 47.72-100 percentage points across seven perturbation groups; guidance has smaller, model-dependent effects. With identical training tasks, decision-record counts, and updates, perturbation-enriched judge training improves ACPBench-Hard success from 35.38% to 42.12% (+6.73 points). SWE verifier training gains 7.60-9.40 points over interface-matched untrained references. Workflow and post-training evaluations preserve task-solving topology with unperturbed inputs and no private-reference guidance. These results support designing collection conditions to obtain corrective supervision that improves collaborative decisions with frozen solving agents.

cs.LG↗

arXiv2026

ThermE: Predictive Management of Shared Thermal Headroom for Sustained LLM Inference on Thermally Constrained Edge SoCs

Compact edge system-on-chip (SoC) platforms increasingly run sustained LLM inference under thermal constraints, while their CPU, GPU, and RAM share a cooling path. Prefill and decode therefore consume shared, time-varying thermal headroom, yet vendor governors react only near hardware throttling thresholds without knowledge of request state or upcoming work. A control decision that improves current performance can thus consume headroom too quickly and degrade subsequent service. In this paper, we present ThermE, a runtime system that predicts and jointly manages shared thermal headroom for sustained LLM inference on edge SoCs. Its Fast LLM-to-Heat Compiler maps the model, requests, runtime state, and candidate actions to domain heats without executing LLMs. A partial differential equation (PDE)-constrained Headroom Predictor uses ThermPINN for offline thermal identification and a Reduced Headroom Predictor (RHP) for low-overhead online uncertainty-calibrated headroom forecasts. An Uncertainty-Aware Action Scheduler then selects actions that balance serving quality and future headroom. We implement ThermE atop vLLM and evaluate it across four LLM inference workloads. The results show that ThermE reduces TTFT and TPOT by 40.55% and 12.17%, respectively, relative to vLLM. It achieves a 5.70% SLO violation rate, compared with 12.30% for the strongest baseline, while its predictor obtains a 1.94 $^\circ$C MAE with 11.50 ms overhead.

eess.SY↗

arXiv2026

DevicesWorld: Benchmarking Cross-Device Agents in Heterogeneous Environments

LLM-based agents have rapidly improved at operating individual digital environments such as mobile applications, desktop systems, and smart homes. However, real-world user goals often span multiple devices: information may come from a phone, be processed on a desktop, and the result may need to appear on another device. Most existing benchmarks center on a single dominant execution environment, making it difficult to evaluate whether agents can acquire and integrate information across heterogeneous devices and complete end-to-end tasks with cross-device dependencies. We introduce DevicesWorld, a large-scale executable benchmark for cross-device collaborative operation. DevicesWorld contains 6,140 tasks and integrates three classes of device environments -- mobile, desktop, and IoT -- into a unified cross-device interaction and evaluation framework. Each task defines a natural-language user goal, participating devices and initial states, executable actions, rule-based verifiers, and a cleanup procedure. A multi-stage construction and quality-control pipeline keeps tasks close to realistic user needs while allowing final outcomes to be automatically verified from device states and generated files. We evaluate five frontier LLM-agent systems on a fixed evaluation set. All methods achieve low success rates, with the best reaching only 12.5%. Among failed runs, about 28.7% satisfy at least one scoring condition yet still fail the full task. Trajectories show that agents become stuck acquiring information or manipulating interfaces, confuse source and output devices, or terminate before all conditions are jointly satisfied. DevicesWorld turns cross-device collaborative operation into an executable, reproducible, and diagnostically useful evaluation problem for research on reliable cross-device agents.

cs.CL↗

arXiv2026

Calyx: Privacy-Preserving Multi-Token Optimistic-Rollup Protocol

Rollup protocols have recently received significant attention as a promising class of Layer 2 (L2) scalability solutions. By utilizing the Layer 1 (L1) blockchain solely as a bulletin board for a summary of the executed transactions and state changes, rollups enable secure off-chain execution while avoiding the complexity of other L2 mechanisms. However, to ensure data availability, current rollup protocols require the plaintext of executed transactions to be published on-chain, resulting in inherent privacy limitations. In this paper, we address this problem by introducing Calyx, the first privacy-preserving multi-token optimistic-Rollup protocol. Calyx guarantees full payment privacy for all L2 transactions, revealing no information about the sender, recipient, transferred amount, or token type. The protocol further supports atomic execution of multiple multi-token transactions and introduces a transaction fee scheme to enable broader application scenarios while ensuring the sustainable operation of the protocol. To enforce correctness, Calyx adopts an efficient one-step fraud-proof mechanism. We analyze the security and privacy guarantees of the protocol and provide an implementation and evaluation. Our results show that executing a single transaction costs approximately $0.06 (0.00002 ETH) and incurs only constant-size on-chain cost in asymptotic terms.

cs.CR↗

arXiv2026

PRISM: A Benchmark for Programmatic Spatial-Temporal Reasoning

Programmatic video generation through code offers geometric precision and temporal coherence beyond pixel-level diffusion models, yet rigorously evaluating whether language models can produce spatially correct animated outputs remains an open problem. We introduce PRISM, a large-scale benchmark of 10,372 human-calibrated instruction-code pairs (20 times larger than prior programmatic video generation benchmarks), grounded in real-world knowledge visualization scenarios across English and Chinese and spanning 437 subject categories. We further propose a funnel-style evaluation framework with four complementary metrics: Code-Level Reliability for executability, Spatial Reasoning for layout correctness over full animation sequences, and Prompt-Aware Dynamic Visual Complexity (PADVC) and Temporal Density (TD) for diagnosing dynamic expression and temporal activity. Systematic evaluation of seven mainstream LLMs reveals a striking Execution-Spatial Gap: the average drop from execution success rate to spatial pass rate is approximately 41%, showing that runnable code does not necessarily yield spatially coherent visual output. These findings show that programmatic video generation evaluation should go beyond executability. PRISM provides a principled benchmark for advancing spatially coherent code generation.

cs.AI↗

arXiv2026

A Security Framework for General Blockchain Layer 2 Protocols

Layer 2 (L2) protocols, payment channels, sidechains, and rollups, are central to blockchain scalability, enabling off-chain execution while preserving on-chain security. Despite growing deployment, existing security models remain protocol-specific and monolithic, hindering compositional reasoning and principled comparison of assumptions and requirements. We present a general security framework for L2 protocols in the IITM-style Universal Composability (iUC) model. At its core is a modular ideal functionality F_layer2 that abstracts mechanism-specific details while capturing the essential structure of L2 systems through composable subroutines for joining, submission, updating, reading, and settlement under adversarial conditions. This yields uniform definitions of safety, liveness, and data availability across a broad class of L2 protocols. We demonstrate generality by instantiating the framework for three representative constructions: the Brick payment channel, the Liquid sidechain, and the Arbitrum Nitro rollup. Each case study yields a protocol-specific ideal functionality derived from F_layer2 and tailored to its assumptions. Our analysis (i) establishes security via simulation-based proofs, (ii) exposes inherent trade-offs among safety, liveness, and data availability, and (iii) derives lower bounds characterizing fundamental limitations of each design class. Finally, we illustrate the framework as a design tool by presenting FRoll, the first optimistic rollup protocol with fast-finality guarantees, together with a security analysis in our model, showing how the framework supports requirement-driven design of L2 protocols.

cs.CR↗

arXiv2026

RefDecoder: Enhancing Visual Generation with Conditional Video Decoding

Video generation powers a vast array of downstream applications. However, while the de facto standard, i.e., latent diffusion models, typically employ heavily conditioned denoising networks, their decoders often remain unconditional. We observe that this architectural asymmetry leads to significant loss of detail and inconsistency relative to the input image. To address this, we argue that the decoder requires equal conditioning to preserve structural integrity. We introduce RefDecoder, a reference-conditioned video VAE decoder by injecting high-fidelity reference image signal directly into the decoding process via reference attention. Specifically, a lightweight image encoder maps the reference frame into the detail-rich high-dimensional tokens, which are co-processed with the denoised video latent tokens at each decoder up-sampling stage. We demonstrate consistent improvements across several distinct decoder backbones (e.g., Wan 2.1 and VideoVAE+), achieving up to +2.1dB PSNR over the unconditional baselines on the Inter4K, WebVid, and Large Motion reconstruction benchmarks. Notably, RefDecoder can be directly swapped into existing video generation systems without additional fine-tuning, and we report across-the-board improvements in subject consistency, background consistency, and overall quality scores on the VBench I2V benchmark. Beyond I2V, RefDecoder generalizes well to a wide range of visual generation tasks such as style transfer and video editing refinement.

cs.CV↗

arXiv2026

AgentXRay: White-Boxing Agentic Systems via Workflow Reconstruction

Large Language Models have shown strong capabilities in complex problem solving, yet many agentic systems remain difficult to interpret and control due to opaque internal workflows. While some frameworks offer explicit architectures for collaboration, many deployed agentic systems operate as black boxes to users. We address this by introducing Agentic Workflow Reconstruction (AWR), a new task aiming to synthesize an explicit, interpretable stand-in workflow that approximates a black-box system using only input-output access. We propose AgentXRay, a search-based framework that formulates AWR as a combinatorial optimization problem over discrete agent roles and tool invocations in a chain-structured workflow space. Unlike model distillation, AgentXRay produces editable white-box workflows that match target outputs under an observable, output-based proxy metric, without accessing model parameters. To navigate the vast search space, AgentXRay employs Monte Carlo Tree Search enhanced by a scoring-based Red-Black Pruning mechanism, which dynamically integrates proxy quality with search depth. Experiments across diverse domains demonstrate that AgentXRay achieves higher proxy similarity and reduces token consumption compared to unpruned search, enabling deeper workflow exploration under fixed iteration budgets.

cs.AI↗

arXiv2026

A multi-stage soft computing framework for complex disease modelling and decision support: A liver cirrhosis case study

Liver cirrhosis is a major global health problem causing millions of deaths annually, and timely detection with aggressive treatment can significantly improve patients' quality of life. Modelling complex diseases from biomedical data is computationally challenging due to high dimensionality, strong feature correlations, noise, and limited labelled samples. Conventional Machine Learning (ML) pipelines often struggle with robustness, interpretability, and generalisation under such conditions. In this study, we propose an ML-driven multi-stage decision framework for complex disease modelling and therapeutic exploration. The framework integrates single-cell transcriptomic profiling, high-dimensional network-based feature stabilisation, multi-model learning, deep representation construction, and post-hoc decision support. Specifically, single-cell sequencing data were analysed to identify key cellular subpopulations, followed by high-dimensional weighted gene co-expression network analysis (hdWGCNA) to stabilise gene modules under sparsity and noise. To enhance non-linear feature interaction modelling, tabular molecular features were restructured into two-dimensional disease maps and analysed using a CNN. Finally, molecular docking was incorporated as a decision-support module to evaluate candidate therapeutic compounds. Using liver cirrhosis as a representative case, the framework identified a disease-associated endothelial subpopulation and extracted seven robust signature genes (HSPB1, GADD45A, CLDN5, ATP1B3, C1QBP, ENPP2, and PARL). The CNN-based representation learning module outperformed conventional pipelines in classification. The framework is disease-agnostic and readily extends to other omics-driven biomedical applications involving uncertainty, heterogeneity, and limited samples.

q-bio.OT↗

arXiv2026

TeachMaster: Generative Teaching via Code

The scalability of high-quality online education is hindered by the high costs and slow cycles of manual content creation. Despite advancements in video generation, current approaches often fail to ensure pedagogical structure and precise control due to their pixel-level, black-box nature. In this paper, we propose Generative Teaching, a novel paradigm shifting educators from manual creators to high-level directors who focus on pedagogical intents while agents handle the execution. To realize this vision, we introduce TeachMaster, a multi-agent framework that leverages code as an intermediate semantic medium. Unlike traditional video generation methods, TeachMaster orchestrates a collaborative team of agents, spanning planning, design, and rendering, to automate the production of interpretable, editable, and curriculum-ready educational videos. Experiments validate that TeachMaster significantly boosts production efficiency without compromising structural coherence or visual fidelity, slashing production costs to only 0.3% of traditional online course videos and providing a robust solution for scalable education.

cs.CY↗

arXiv2026

Composed Vision-Language Retrieval for Skin Cancer Case Search via Joint Alignment of Global and Local Representations

Medical image retrieval aims to identify clinically relevant lesion cases to support diagnostic decision making, education, and quality control. In practice, retrieval queries often combine a reference lesion image with textual descriptors such as dermoscopic features. We study composed vision-language retrieval for skin cancer, where each query consists of an image to text pair and the database contains biopsy-confirmed, multi-class disease cases. We propose a transformer based framework that learns hierarchical composed query representations and performs joint global-local alignment between queries and candidate images. Local alignment aggregates discriminative regions via multiple spatial attention masks, while global alignment provides holistic semantic supervision. The final similarity is computed through a convex, domain-informed weighting that emphasizes clinically salient local evidence while preserving global consistency. Experiments on the public Derm7pt dataset demonstrate consistent improvements over state-of-the-art methods. The proposed framework enables efficient access to relevant medical records and supports practical clinical deployment.

cs.CV↗

arXiv2026

TinyGaze: Lightweight Gaze-Gesture Recognition on Commodity Mobile Devices

Gaze gestures can provide hands free input on mobile devices, but practical use requires (i) gestures users can learn and recall and (ii) recognition models that are efficient enough for on-device deployment. We present an end-to-end pipeline using commodity ARKit head/eye transforms and a scaffolded guidance-to-recall protocol grounded in learning theory. In a pilot feasibility study (N=4 participants; 240 trials; controlled single-session setting), we benchmark a compact time-series model (TinyHAR) against deeper baselines (DeepConvLSTM, SA-HAR) on 5-way gesture recognition and 4-way user identification. TinyHAR achieves strong performance in this pilot benchmark (Macro F1 = 0.960 for gesture recognition; Macro F1 = 0.997 for user identification) while using only 46k parameters. A modality analysis further indicates that head pose dynamics are highly informative for mobile gaze gestures, highlighting embodied head--eye coordination as a key design consideration. Although the small sample size and controlled setting limit generalizability, these results indicate a potential direction for further investigation into on-device gaze gesture recognition.

cs.HC↗

arXiv2026

GazeSync: A Mobile Eye-Tracking Tool for Analyzing Visual Attention on Dynamically Manipulated Content

Conventional mobile eye-tracking maps gaze to static screen coordinates, failing to capture user attention when content is dynamic. As users pinch, zoom, and rotate images, static coordinates lose their semantic meaning relative to the underlying visual content. To address this methodological gap, we present \textit{GazeSync}, a reusable mobile system that synchronizes on-device gaze estimation with real-time image transformation matrices (scale, rotation, and translation). By logging gaze coordinates alongside precise UI states, GazeSync enables the accurate reconstruction of \textit{image-relative} attention patterns, decoupling visual attention from device interaction. We validate our end-to-end toolchain through a formative study involving guided manipulation, reading, and visual search tasks. Our results demonstrate GazeSync's ability to recover ground-truth gaze locations on transforming content, explicitly showing how it outperforms static baselines, while also surfacing critical boundaries regarding calibration drift and reconstruction fragility under compound manipulations.

cs.HC↗

arXiv2026

Towards a Science of Collective AI: LLM-based Multi-Agent Systems Need a Transition from Blind Trial-and-Error to Rigorous Science

Recent advancements in Large Language Models (LLMs) have greatly extended the capabilities of Multi-Agent Systems (MAS), demonstrating significant effectiveness across a wide range of complex and open-ended domains. However, despite this rapid progress, the field still relies heavily on empirical trial-and-error. It lacks a unified and principled scientific framework necessary for systematic optimization and improvement. This bottleneck stems from the ambiguity of attribution: first, the absence of a structured taxonomy of factors leaves researchers restricted to unguided adjustments; second, the lack of a unified metric fails to distinguish genuine collaboration gain from mere resource accumulation. In this paper, we advocate for a transition to design science through an integrated framework. We advocate to establish the collaboration gain metric ($Γ$) as the scientific standard to isolate intrinsic gains from increased budgets. Leveraging $Γ$, we propose a factor attribution paradigm to systematically identify collaboration-driving factors. To support this, we construct a systematic MAS factor library, structuring the design space into control-level presets and information-level dynamics. Ultimately, this framework facilitates the transition from blind experimentation to rigorous science, paving the way towards a true science of Collective AI.

cs.CL↗