Search arXiv⌕ Search

arXiv subjects

Ya Zhang

Publications and source records attributed to Ya Zhang.

At least 19 recordsLinked to original sources
arXiv2026

Advancing Clinical Genomic Analysis with a Skill-Centred Agent

Clinical genomic analysis requires experts to integrate evolving guidelines, databases, and computational evidence. As sequencing has become increasingly accessible, this complex and labor-intensive process has become a major bottleneck to the timely use of genomic information in clinical decision-making. To fill this gap, we developed ClinGeX, a large-language-model agent that composes 105 curated executable skills through task-specific retrieval and dependency-aware orchestration. On 223 temporally held-out benchmark cases, ClinGeX achieved 71.7\% macro-averaged task-level concordance with expert reference classifications, compared with 45.7\% for the strongest comparator. In a three-condition, counterbalanced human-AI collaboration study, 45 clinical genomic professionals assessed 135 clinically derived cases under routine (no AI assistance), DeepSeek-assisted and ClinGeX-assisted conditions. ClinGeX assistance increased raw case-level concordance across criterion-level variant analysis, genomic interpretation in rare disease and molecular oncology decision-making (51.1\%, 57.8\%, 64.4\% versus 15.6\%, 31.1\%, 20.0\% in comparator conditions) and reduced reported interpretation time in all scenarios. During an exploratory one-month real-world non-interventional assessment, 11 experienced professionals submitted feedback on 96 self-selected analysis episodes and reported frequent conclusion adoption and time savings. These findings support executable domain skills as a promising approach to augment specialist genomic interpretation.

cs.AI↗
arXiv2026

Kinetic Sectors and Nonequilibrium Dynamics in the $δ$-Ising Model

We introduce and analyze the $δ$-Ising model, a kinetic Ising ferromagnet in which single-spin flips are accepted when their energy change does not exceed a tunable threshold. Varying the field and threshold opens or closes local flip channels, dividing parameter space into kinetic sectors, each governed by a fixed Markov transition operator. We prove that the dynamics retains global spin-inversion symmetry exactly when up and down spins have equal numbers of open channels. In one such sector on a periodic chain of length $L$, unbiased random initial states therefore reach the field-aligned and antifield states with probability $1/2$ each. Closing a single channel leaves both states absorbing but makes the probability of ordering against the field decay as $(2/π)^L$. Parameter sweeps correspond to ordered products of sector operators, which separate reaching a state from holding it, so hysteresis appears when a state is reached and released at different sector boundaries. We also show how increasing the system size can induce a Mpemba-like reversal, allowing a state with more misaligned spins to reach the final ordered state sooner on average. On a periodic chain, this reversal arises because the mean absorption time grows logarithmically with system size for many short domains but linearly for a single macroscopic domain, and it disappears when the bulk flip channel is opened. These findings show how a simple, intuitive rule for Ising dynamics links local kinetic constraints to field response, memory, and relaxation within a unified framework.

cond-mat.stat-mech↗
arXiv2026

Asymmetric Looped Transformers: Fixed Global Representations for an Iterative Decoder

Looped Transformers increase effective depth by repeatedly applying shared layers, offering a way to scale computation without increasing parameter count. However, full-stack looping places context construction and iterative state refinement within the same recurrent computation, without explicitly distinguishing their roles. We introduce the Asymmetric Looped Transformer (AsymLoop), which separates single-pass context encoding from recurrent state refinement. A single-pass causal encoder produces layer-specific Global key-value (KV) caches that remain fixed across iterations, while a weight-tied recurrent decoder generates Local KV from its evolving state. Joint attention over both sources under a single softmax allows each update to integrate persistent contextual information with the current intermediate state. For training, we apply a heavy-tail-guided layerwise learning-rate method to the encoder and decoder. Experiments on FineWeb-Edu across four model scales, from 60M to 1B, show consistently lower validation perplexity and higher average downstream accuracy than Full Loop under matched effective depth and training token budgets. At 1B, AsymLoop improves average zero-shot accuracy across seven commonsense reasoning benchmarks from 49.58% to 50.85%. Further experiments extend these benefits to longer training budgets, an OLMo-style backbone, and sparse mixture-of-experts models. These results support asymmetric context encoding and state refinement as an effective design for recurrent language models.

cs.CL↗
arXiv2026

World Action Modeling with Progressive Visual Planning

World action models (WAMs) have emerged as a promising paradigm for robotic control by jointly predicting future visual dynamics and actions from an initial observation and instruction. However, existing WAMs struggle with long-horizon prediction, as generating dense video rollouts is highly inefficient. Some recent WAMs address this by predicting a single future frame without generating the full video, but this approach neglects how to progress toward the goal. We present ProWAM, a progressive world action model that jointly predicts actions and an ordered sequence of sparse visual sub-goals, providing explicit visual guidance to anchor action generation throughout task execution. This design scales naturally, as sub-goal prediction can be learned from large-scale action-free videos, allowing the video backbone to offload complex visual planning from the action policy. For efficient action generation, ProWAM executes a single video-backbone forward pass to cache sparse sub-goal features, eliminating iterative full-video generation and requiring only lightweight action denoising during replanning. Across extensive evaluations, ProWAM achieves superior out-of-distribution robustness. On simulation benchmarks, it sets new state-of-the-art results on LIBERO-Plus (85.8%) and randomized RoboTwin (75.7%), outperforming the strongest baseline with relative gains of up to +35.9%. On RoboCasa365, ProWAM achieves a 48.1% success rate and 18.2% on the challenging Composite-Unseen split, ranking 4th overall. Crucially, in zero-shot real-world experiments, ProWAM achieves 70.0% success, outperforming the strongest baseline by +15.0 (from 55.0% to 70.0%, a +27.3% relative gain) in novel scenes. These results demonstrate the value of progress-indexed visual foresight for closed-loop control. Our program is in https://sii-ferenas.github.io/ProWAM-page.

cs.AI↗
arXiv2026

Q-learning Penalized Transformer for Safe Offline Reinforcement Learning

This paper addresses the problem of safe offline reinforcement learning, which involves training a policy to satisfy safety constraints using an offline dataset. This problem is inherently challenging as it requires balancing three highly interconnected and competing objectives: satisfying safety constraints, maximizing rewards, and adhering to the behavior regularization imposed by the offline dataset. To tackle this trilogy challenge, we propose Q-learning Penalized Transformer policy (QPT), a \emph{training--inference consistent} framework that bridges conditional sequence modeling with constraint-aware value estimation. QPT trains a Transformer policy that generates actions conditioned on trajectory context and target return/cost, retaining strong behavior regularization. To inject explicit safety semantics during learning, we augment sequence-model training with a Q-shaped penalty using learned reward and cost Q-functions to favor high return under low constraint violation. At inference, the same Q-functions enforce the cost threshold and choose the highest-reward feasible action, closing the loop between training and deployment. We provide a principled analysis under stylized near-deterministic CMDPs, characterizing how Q-penalized conditional generation improve safety and performance. Empirically, QPT consistently outperforms strong safe offline RL baselines across 38 tasks on the DSRL benchmark, and exhibits robust zero-shot adaptation to different constraint thresholds.

cs.LG↗
arXiv2026

Alignment-Guided Flow Transformer for Efficient Vision-Language-Action Policy Learning

Recent advances in Vision-Language-Action (VLA) models point toward general-purpose robotic intelligence by unifying perception, instruction, and control. Despite impressive progress, existing VLA models often adapt poorly due to \emph{tri-modal misalignment} among vision, language, and action, which weakens action grounding and hurts generalization and fine-tuning efficiency. In this work, we present Alignment-Guided Flow Transformer (AGFT), a novel framework that explicitly enforces tri-modal alignment through a dedicated alignment loss, bridging the representational gap across modalities and enhancing task adaptation. While prior research has predominantly emphasized bi-modal vision--language alignment, we systematically formalize and study tri-modal alignment in VLA models, and provide both ablations and analysis to isolate its role in improving adaptation and robustness. To further accelerate deployment, we adopt a flow-matching objective, enabling substantially fewer inference steps than diffusion-based policies while maintaining accuracy. Theoretically, we establish a quantitative connection between the tri-modal alignment gap and the optimization tightness of flow matching; empirically, experiments on the extensive benchmark show that AGFT achieves superior success rates and lower inference latency compared to SOTA baselines, underscoring tri-modal alignment as a key ingredient for scaling robust VLA manipulation.

cs.LG↗
arXiv2026

SkillCome: Group Contrast Skill Optimization with Dual Memory

Skill evolution improves the capabilities of large language models by analyzing trajectories generated under a given skill and modifying the skill accordingly. Existing approaches typically generate a single trajectory per question. However, this provides insufficient optimization signals since it requires inferring effective skill edits from a solitary path. It is difficult to pinpoint which actions caused the failure in a failed trajectory, or to determine which actions in a successful one should be incorporated into the skill. Furthermore, they rely on a local batch of trajectories for analysis, making the optimization direction susceptible to noisy evidence. To address these, we propose SkillCome, a Skill-evolution method based on group Contrast optimization with dual memory. For each question, SkillCome generates trajectories and performs group contrast analysis to precisely identify key behavioral divergences between successful and failed trajectories, offering reliable optimization signals. The dual memory system further accumulates evidence from historical steps to track patterns shared across different groups, leading to more generalized optimization directions. Together, SkillCome builds a systematic optimization process that transforms experience from observed successful trajectories into reusable skills. Extensive experiments on six benchmarks spanning question answering, reasoning, and agentic tasks demonstrate the effectiveness of our method. SkillCome consistently outperforms baselines across five models of varying families and scales, with gains up to +5.69 points.

cs.AI↗
arXiv2026

Phasor-particle-in-cell algorithm for bidirectional external-circuit coupling of inductively coupled plasma

We develop a self-consistent phasor-particle-in-cell algorithm to capture bidirectional inductive and capacitive feedback between an inductively coupled plasma and a distributed radio-frequency coil circuit. The method couples a two-dimensional axisymmetric particle-in-cell model with Monte Carlo collisions to a network that resolves the current and potential of each coil turn. Precomputed unit-current Helmholtz solutions provide the vacuum impedance matrix, while projection of the plasma-only field supplies the series back electromotive force. An established charge-based electrostatic coupling is extended to individual turns: surface charges obtained from a discrete Gauss law consistent with the Poisson solver yield shunt displacement currents through their fundamental harmonics. Both responses are accumulated over a complete radio-frequency period to update the relaxed circuit state once per period. Vacuum tests showed differences of -1.58 percent in single-solenoid inductance and -4.38 percent in mutual inductance relative to analytical references. A particle-free five-node manufactured benchmark with both coupling channels active yielded a scaled complex-state error of 1.90 x 10^-5 against the continuous reference and approximately second-order spatial convergence. Application to an unshielded argon reference cell reached statistically stationary states at coil-port powers of 60 and 100 W. Net capacitive heating accounted for 29.35 percent and 20.07 percent, respectively, of the combined inductive and capacitive plasma power. Independently accumulated field-side and circuit-side powers, including conductor loss, agreed within approximately 0.5 percent. These results support turn-resolved bidirectional field-circuit coupling for axisymmetric plasma simulations using a fundamental-frequency circuit representation.

physics.plasm-ph↗
arXiv2026

How Well Can Modern LLMs Act as Agent Cores in Radiology Environments?

Radiology, with its heterogeneous modalities, anatomies, and evolving protocols, is a natural yet high-stakes testbed for agentic AI. As LLMs grow more capable and tool ecosystems such as MCP and Agent Skills more complex, their capability boundaries in radiology remain unclear. We introduce RadA-BenchPlat, a two-layer benchmark spanning idealized reasoning and real-world execution: a synthetic layer with 2.2k clinician-verified records, 24.2k QA pairs, and 10 tool categories under diverse availability settings; and a real-environment layer pairing 165 2D/3D cases with executable tools for grounding, diagnosis, and report generation. We find that high completion rates under idealized settings do not carry over to real environments, where the best agents trail their upper bound by approximately 15% and drop to 0.327 on long-chain tasks. Prompting strategies (few-shot exemplars, multi-agent coordination) and AutoTB (on-the-fly synthesis of missing tools) mitigate execution drift and breakdowns in difficult scenarios, lifting execution success to 94.5%, yet remain not fully reliable. Overall, our findings suggest that radiology agents are no longer limited to conceptual demonstrations, but are beginning to show potential for real-world applications. Project repository: https://github.com/MAGIC-AI4Med/RadABench

cs.CV↗
arXiv2026

ChatBEV: Empowering Traffic Scene Understanding and Simulation via Vision-Language Model

Comprehensive traffic scene understanding is a foundational capability for Intelligent Transportation Systems (ITS) underpinning applications such as traffic simulation. While VisionLanguage Models (VLMs) have demonstrated strong reasoning potential, their application to Bird's-Eye View (BEV) maps in traffic contexts remains limited by narrow task definitions and scarce annotated data. We introduce ChatBEV-QA, a large-scale BEV VQA benchmark of 137K+ QA pairs, designed to evaluate global scene understanding, vehicle-lane interactions, and vehiclevehicle interactions within complex traffic environments. Building on this, we fine-tune ChatBEV, a specialized VLM that accurately interprets diverse scene understanding queries from BEV maps. To demonstrate downstream utility in ITS applications, we integrate ChatBEV into a language-guided traffic simulation framework. Its global understanding and navigation reasoning provide crucial context-aware guidance, reducing trajectory displacement error by up to 20.9% and scenario collision rates by up to 37.9% over text-only baselines.

cs.CV↗
arXiv2026

GLoRI: Closed-Loop Whole-Body Tracking with Global-Local Reference Interaction for Humanoid Loco-Manipulation

Humanoid loco-manipulation requires accurate whole-body motion tracking in the world frame for physical interaction. While local references preserve motion structure, they lack explicit constraints on absolute spatial placement, leading to accumulated global errors. Existing globally aware approaches augment teleoperation policies with global observations but do not explicitly integrate global correction with local motion guidance, limiting autonomous tracking accuracy. We present GLoRI, a closed-loop whole-body controller that integrates structured global reference and feedback with local motion guidance. Its GLoRI-Net uses Global-Local Cross Attention(GLCA) to refine local keypoint features with global target and pose-difference features, preserving motion structure while correcting world-frame placement. GLoRI achieves 100% completion and a g-MPJPE of 6.44cm on held-out HuMoTo motions. This accuracy remains robust under direct Isaac Gym-to-MuJoCo transfer without fine-tuning, demonstrating strong generalization. Furthermore, such accuracy and generalization enable autonomous loco-manipulation with a single policy on a real Unitree G1 interacting with diverse unseen objects, extending beyond prior systems that primarily rely on teleoperation or focus on single-object interactions.

cs.RO↗
arXiv2026

Multi-Source Retrieval and Reasoning for Legal Sentencing Prediction

Legal judgment prediction (LJP) aims to predict judicial outcomes from case facts and typically includes law article, charge, and sentencing prediction. While recent methods perform well on the first two subtasks, legal sentencing prediction (LSP) remains difficult due to its need for fine-grained objective knowledge and flexible subjective reasoning. To address these limitations, we propose $MSR^2$, a framework that integrates multi-source retrieval and reasoning in LLMs with reinforcement learning. $MSR^2$ enables LLMs to perform multi-source retrieval based on reasoning needs and applies a process-level reward to guide intermediate subjective reasoning steps. Experiments on two real-world datasets show that $MSR^2$ improves both accuracy and interpretability in LSP, providing a promising step toward practical legal AI. Our code is available at https://github.com/cjj826/MSR2.

cs.IR↗
arXiv2026

Counter-rotating density and current structures in a partially magnetized $\mathbf{E}\times\mathbf{B}$ plasma

The first 3D kinetic simulation of a magnetron discharge reproduces the measured rotating spoke and reveals that density and current rotate in opposite directions: the $m=1$ spoke turns in the $\mathbf{E}\times\mathbf{B}$ direction at 90~kHz while $m\simeq16$ electron-cyclotron-drift-type filaments turn at 1.0~MHz, with no net propagation along $\mathbf{B}$. Magnetic drifts exchange far more energy than they deposit, yet their spoke-front heating sustains the ionization. Following helical paths, the anode-directed current is relayed by de-trapping turbulence at the sheath edge (50\%) and by the spoke channel in the bulk (97\%).

physics.plasm-ph↗
arXiv2026

SpotSound: Enhancing Large Audio-Language Models with Fine-Grained Temporal Grounding

Large Audio-Language Models (ALMs) have recently demonstrated remarkable capabilities in holistic audio understanding, yet they remain unreliable for temporal grounding, i.e., the task of pinpointing exactly when an event occurs within long-form audio. This limitation stems from two factors: training data dominated by clip-level supervision lacking precise timestamps, and benchmarks that fail to simulate real-world scenarios where short events are obscured by dense background sounds. In this paper, we introduce SpotSound, an audio language model designed for grounding audio events. SpotSound incorporates a novel training objective, specifically designed to suppress hallucinated timestamps for events absent from the input. Additionally, we present SpotSound-Bench, a challenging temporal grounding benchmark where target events occupy less than ~10% of each clip, creating a rigorous 'needle-in-a-haystack' evaluation. Experiments demonstrate that SpotSound achieves state-of-the-art results on temporal grounding benchmarks while maintaining robust performance across general downstream audio-language tasks. Code, models and benchmark are released on https://loiesun.github.io/spotsound/

cs.SD↗
arXiv2026

FailForge: Distilling Procedural Competence from Persistent Failures into Code Agents

Rejection sampling fine-tuning (RFT) is widely used to train code agents by generating trajectories on verifiable software engineering tasks, retaining those that pass the tests, and fine-tuning on the successful rollouts. However, even strong code agents repeatedly fail on a substantial fraction of such tasks, and standard RFT simply discards these failures. The discarded samples are precisely the hardest and most informative ones, drawn from verifiable instances that are costly to curate. Stronger base models may reduce the number of failures, but the remaining hard cases still define the frontier for further improvement. We propose FailForge, an agentic framework that converts failed rollouts into training signal. For each failed instance, an agent diagnoses the failure from error feedback and execution traces, distills the diagnosis into a concise and actionable skill, and injects the skill into the agent context for a guided second attempt. Trajectories that succeed under skill guidance are folded back into the RFT corpus. Crucially, the skill is removed at training time, so the model internalizes the recovered behavior rather than relying on external hints at inference. FailForge recovers over 26% of previously failed instances at marginal additional cost, and training Qwen3.5-4B on the augmented corpus improves the SWE-bench Verified resolve rate by 6.6 points over a strong RFT baseline, with gains concentrated on the hardest problems.

cs.AI↗
arXiv2026

Modular TTT: Rethinking Test-Time Training as Composable Modules

Test-time training (TTT) views sequence modeling as an online learning problem in which fast weights are updated by an internal learning rule. Despite the growing number of TTT variants, existing approaches typically hard-code each variant separately, which makes it difficult to design new TTT methods and to isolate the role of each component. To address this, we propose Modular TTT, a framework that represents the inner learner as a directed acyclic graph and exposes the fast-weight network, loss function, learning rate, weight decay, and normalization as explicit design dimensions. Modular TTT automatically composes primitive-level train-view forward, train-view backward, and causal query-view rules into the full graph-level TTT computation, including the fast-weight state transition. Using Modular TTT, we systematically ablate the components of TTT and find that small learning-rate initialization, weight decay, and a single-layer nonlinearity improve performance, while MSE and inner-product losses perform similarly. Deeper fast-weight networks and normalization tend to hurt performance because they induce excessively large activations, while residual connections and gating provide little measurable benefit. Guided by these findings, we train the best resulting variant as 410M- and 1.45B-parameter models on 100B tokens, and observe training loss and benchmark performance comparable to Gated DeltaNet.

cs.LG↗
arXiv2026

CW-Ghost: Search-Free Granularity Selection for Helper-Thread Prefetching via Capacity Windows

Helper-thread prefetching hides the latency of irregular memory accesses by executing address dependency chains ahead of the main thread. However, its effectiveness depends on the range of future iterations covered by the helper thread. A fixed coverage range cannot consistently accommodate different workloads and processors, whereas exhaustively evaluating candidate configurations incurs substantial configuration cost. This paper presents CW-Ghost, which uses a single offline profiling run to estimate the average demand cache line fill volume generated per target iteration in a target region. CW-Ghost combines this estimate with a cache capacity budget to derive a Capacity Window, which determines the iteration granularity of each prefetch chunk. In addition, bounded chunk-level synchronization limits the number of chunks by which the helper thread may run ahead of the main thread. Across 14 workload instances evaluated on Intel and AMD CPU platforms, CW-Ghost achieves geometric mean speedups of 1.54x and 1.33x, respectively, over the original programs. Compared with Ghost Threading, it improves geometric mean performance by 15.8% and 10.8%, respectively, while achieving more than 99% of the empirically optimal performance within the candidate set on both platforms. These results demonstrate that cache capacity constraints can effectively guide the selection of granularity for helper-thread prefetching.

cs.DC↗
arXiv2026

Dementia Etiology Diagnosis via Collaborative Meta Knowledge Enhancement

Although artificial intelligence (AI) has shown promising performance in several medical tasks, accurate dementia etiology diagnosis with AI remains challenging due to complex overlapping symptoms among diseases. Scaling up the dataset size by combining the cross-center samples may bring a gain in the pursuit of performance, while the inherent data heterogeneity across centers or populations induces the conflict. Conventional multi-task learning paradigms offer a promising framework; however, they fail to consider critical meta information (e.g., site-specific acquisition and modality availability) to combat the heterogeneity. To address this challenge, we propose a Collaborative Meta Knowledge Enhancement (COME) framework for dementia etiology diagnosis, which injects multi-center acquisition semantics, source identifiers, and modality indicators as heterogeneity-aware embeddings into a unified Transformer architecture for scale-up training, enabling explicit modeling of heterogeneity. Besides, a trust-region constrained optimization scheme is designed to regularize the model from spurious correlations during training through a reference model. Across seven independent cohorts, our method achieves state-of-the-art in-domain performance with a mean macro-averaged AUC of 85.62% and a 4.29-point gain over the strongest baseline, while maintaining superior out-of-domain generalization under both cross-center and cross-sequence evaluations. Extensive validation also confirms the alignment between model predictions and established biomarkers (amyloid, tau) and clinical severity, highlighting the potential of COME to enable robust and interpretable dementia diagnostics in real-world settings.

cs.LG↗