arXiv subjects

Yang Zhou

Publications and source records attributed to Yang Zhou.

At least 37 records · Page 2Linked to original sources

arXiv2026

A Unified Vision-Language Model for PSMA PET/CT Report Generation, Visual Question Answering, and Lesion Segmentation

Accurate PSMA PET/CT interpretation is central to prostate cancer management, yet existing PET/CT AI models typically address isolated tasks. We propose a unified PSMA PET/CT vision-language model for report generation, visual question answering, and lesion segmentation. The framework adopts an LLaVA-style architecture, comprising a PET/CT vision encoder, an MLP-Mixer projection module, a LoRA-tuned large language model, and a 3D segmentation branch. Training followed a four-stage strategy: vision encoder pretraining, projection-layer alignment, VLM fine-tuning, and final multitask tuning. Language tasks used 5,747 PSMA PET/CT datasets with paired reports, while segmentation used the PSMA subset of AutoPET. The model outperformed PET2REP and a CT-based baseline across standard report-generation metrics, improved performance across VQA question types, and achieved higher Dice and lesion-level overlap F1 than SegAnyPET and nnUNet. These results support the feasibility of a unified framework for structured, interactive, interpretable PSMA PET/CT analysis with voxel-level grounding within a single multitask model architecture.

cs.CV↗

arXiv2026

mKernel: Fast Multi-GPU, Multi-Node Fused Kernels

Communication has become a bottleneck in distributed training and inference of large models. Overlapping communication with computation at the granularity of kernels, on separate streams, reduces only part of this communication cost. Fused kernels often have better performance by transmitting each output tile as soon as it is produced, but existing fused kernels are largely confined to a single NVLink domain. We present mKernel, a library of multi-GPU, multi-node fused kernels that overlap computation, intra-node NVLink communication, and inter-node RDMA at tile granularity. mKernel partitions the streaming multiprocessors (SMs) of a persistent kernel into compute and communication roles, and an on-GPU controller tunes the SM partition adaptively at run time, since the best SM partition varies with the kernel and the input shape. It structures data movement hierarchically so that data traversing the inter-node network is minimized. Finally, it drives the network from the GPU through a lightweight command queue and host proxy implemented directly on RDMA verbs, which allows the same kernels to run on any network backend (e.g. InfiniBand and on AWS EFA); we observe, surprisingly, that GPUDirect Async (IBGDA) yields little additional benefit over host-assisted GPU-initiated communication. We implement five kernels spanning tensor, sequence, and expert parallelism. On two 16-GPU H200 clusters, mKernel achieves speedups of up to 1.72x on GEMM+AllReduce and $1.88\times$ on Ring Attention.

cs.DC↗

arXiv2026

Singular Extremal Solutions on Thin Ellipsoids with Varying Nonlinearities

Let $N=m+1$ and consider the thin ellipsoid \[ Ω_\varepsilon=\{(y,x_N)\in\mathbb R^m\times\mathbb R:\ |y|^2+\varepsilon^{-2}x_N^2<1\}. \] We prove that in every sufficiently large dimension, for every sufficiently small $\varepsilon>0$, there exists a smooth positive, strictly increasing, strictly convex, superlinear nonlinearity $f_\varepsilon$ for which the extremal solution in $Ω_\varepsilon$ is an unbounded $H^1_0$ solution. Combining this result with Dancer's thin-domain regularity theorem for the Gelfand nonlinearity $f(t)=e^t$, we obtain on the same sufficiently thin ellipsoids a bounded Gelfand extremal solution and an unbounded extremal solution for another nonlinearity. Thus, in this two-part sense, Brézis' Open Problem~6.1 is resolved in every sufficiently large dimension.

math.AP↗

arXiv2026

Weather-Conditioned Depth Anything

Monocular depth estimation foundation models, such as the Depth Anything series, have achieved remarkable performance across diverse domains. However, they still suffer from critical failures under adverse weather conditions, such as fog, rain, snow, or at night. To address this, we present Weather-Conditioned Depth Anything (DA-W), a framework that explicitly disentangles style from content for weather-robust depth estimation. Specifically, we introduce a Style Filter trained on a curated mix of real and synthetic degradation datasets to extract content-independent, degradation-aware weather embeddings. This style embedding is then injected into the Depth Anything backbone using a parameter-efficient, zero-initialized adapter. Such a lightweight modulation allows a single unified model to robustly adapt to diverse conditions, including fog, rain, snow, and low-light, while avoiding catastrophic forgetting of its core generalization abilities in normal conditions. We train the adapter using a pseudo-label distillation and alignment strategy. Our comprehensive experiments demonstrate that our proposed DA-W achieves state-of-the-art robust depth estimation, improving AbsRel by an average of 3.7% on our curated weather benchmarks, while matching or slightly outperforming performance on standard clean benchmarks. Our project page is available at https://zhaoming-tamu.github.io/WCDA/.

cs.CV↗

arXiv2026

Gradient-Aligned Pair Selection for Personalized Preference Optimization

Personalizing large language models (LLMs) requires aligning generation behavior with user-specific preferences rather than aggregate quality. While Direct Preference Optimization (DPO) provides a stable framework for preference learning, its effectiveness in personalized settings critically depends on how preference pairs are selected. Existing approaches typically rely on heuristic criteria, such as likelihood-based extremes, which decouple optimization from explicit user utility and can lead to degraded personalization. We formalize personalized preference learning as a geometry-aligned optimization problem by analyzing the first-order interaction between gradients of expected user utility and DPO update directions. Our analysis reveals that, under off-policy sampling, the DPO update transitions from a purely error-corrective signal to a reinforcement-like update when preference margins are directionally aligned with utility gradients. This perspective exposes pair selection as a geometric decision that governs whether preference optimization advances or hinders personalization. Motivated by this insight, we propose GAP-DPO (Geometry-Aligned Preference DPO), an iterative algorithm that performs utility-aware, geometry-aligned pair selection while controlling distribution shift via epoch-wise regeneration. Experiments on personalized text generation benchmarks show that GAP-DPO consistently improves stylistic fidelity, preference alignment, and generation quality compared to standard DPO variants. Together, our results establish gradient alignment as a unifying principle for personalized preference optimization and demonstrate that pair selection is an intrinsic component of the optimization geometry rather than a heuristic preprocessing step.

cs.AI↗

arXiv2026

Act More, Decide Less: Skill-Guided Adaptive Action Chunking for Long-Horizon LLM Agents

Large language model (LLM) agents for long-horizon interactive tasks typically follow a ReAct-style protocol, issuing one primitive action per LLM round. While this enables frequent replanning, it is inefficient for long-horizon tasks where many rounds are spent on routine action sequences. A natural alternative is to let the agent emit variable-length action chunks. However, naively training such policies with standard reinforcement learning fails: the agent either collapses to single-action behavior or over-commits to excessively long sequences. Both failures share a common root cause: the inability to learn chunk boundaries. We propose SPACE, which addresses this challenge by distilling chunk-boundary supervision from trajectory-induced programmatic skills. We induce two-level programmatic skills from successful trajectories, where subskill boundaries serve as direct chunk-boundary supervision. This temporal structure is then distilled into a primitive-chunk policy via hybrid on-/off-policy optimization with chunk-aware credit assignment. Experiments on ALFWorld and ScienceWorld show that SPACE improves success rates by 7.0%-31.3% over the strongest baseline in each setting while reducing average LLM decision rounds by up to 78.9%.

cs.LG↗

arXiv2026

OISD: On-Policy Internal Self-Distillation of Language Models

Recent reinforcement learning (RL) post-training approaches primarily optimize the final output policy using sparse outcome-level rewards, while largely overlooking predictive signals encoded in intermediate representations. In this paper, we introduce a new paradigm called on-policy internal self-distillation and propose the OISD framework, which improves reasoning by transferring on-policy predictive signals from the final layer to intermediate representations. During rollout and Group Relative Policy Optimization (GRPO) optimization, the final layer acts as both the policy and a detached internal teacher for selected intermediate layers, which are guided to align with it through two complementary mechanisms: logit alignment, which transfers high-level reasoning behaviors (how to think), and attention alignment, which enforces consistent attention patterns (where to look) from the final layer to the selected intermediate layer, both without requiring external privileged information. Our OISD, together with GRPO, employs signed advantage-weighted Jensen--Shannon alignment to distill informative intermediate representations while preserving policy consistency under a unified acting policy. Experimental results demonstrate the effectiveness of OISD, with substantial and consistent improvements over strong reasoning RL baselines across four mathematical reasoning tasks. The code will be released at https://github.com/THE-MALT-LAB/OISD

cs.LG↗

arXiv2026

Wall-crossing formula and genus-one Virasoro conjecture for Fano complete intersections

The Virasoro conjecture predicts a set of universal relations among all genera Gromov--Witten invariants of any smooth projective variety. The conjecture is well understood for semisimple theories, but remains largely open in the non-semisimple setting. We prove the genus-one Virasoro conjecture on the ambient state space of smooth Fano complete intersections in projective space. For most of these complete intersections, the big quantum cohomology is nowhere semisimple. We also generalize the wall-crossing formula for quasimap invariants with weighted markings to the equivariant twisted setting, allowing descendant insertions at light markings. Together with genus-one quantum Lefschetz for quasimaps with light markings, this wall-crossing formula provides the key bridge from the Gromov--Witten theory of the complete intersection to the semisimple equivariant twisted theory of the projective space.

math.AG↗

arXiv2026

Sharp Gradient Stability for the Sobolev Trace Inequality

Let $n\ge3$ and $1<p<n$. We prove a quantitative stability estimate for the critical Sobolev trace inequality on the upper half-space. More precisely, the Sobolev trace deficit controls the $\max\{2,p\}$-th power of the gradient distance to the manifold of trace bubbles. A central part of the proof is the spectral nondegeneracy of the trace bubbles: the first two eigenspaces of the linearized weighted Steklov problem are exactly the amplitude, dilation, and tangential translation modes.

math.AP↗

arXiv2026

Sharp One-bubble Critical-Point Stability and Global Compactness for the Sobolev Trace Inequality

Let $n\ge3$ and $1<p<n$. We first prove the local trace analogue of the sharp one-bubble critical-point stability theorem of Liu and Zhang~\cite{LiuZhang2025}: near a positive trace-bubble, the Euler--Lagrange residual controls the gradient distance to the normalized trace-bubble manifold with the sharp power $\max\{1,p-1\}$. Then, we establish a Struwe-type compactness theorem for the critical trace functional, which gives the trace counterpart of the Mercuri--Willem decomposition~\cite{MercuriWillem2010}. Combining Struwe-type compactness with the local stability estimate yields a sharp quantitative one-bubble critical-point stability theorem.

math.AP↗

arXiv2026

Two Heads are Better Than One: Test-time Scaling of Multi-agent Collaborative Reasoning

Test-Time Scaling has emerged as a powerful method to extend the reasoning capabilities of Large Language Models. However, single-agent TTS faces significant scalability bottlenecks, as excessively long reasoning traces lead to increased inference costs and stability issues caused by context management failures. To address these limitations, we propose leveraging Multi-Agent Systems as a structural upgrade to standard TTS. By decomposing monolithic reasoning chains into distinct, manageable contexts across multiple agents, MAS offers a more robust framework for scaling reasoning. We validate this approach by introducing M500, a dataset comprising 500 high-quality multi-agent, multi-turn collaborative reasoning traces generated via DeepSeek-R1. Through Supervised Fine-Tuning on M500, we enable open-source models to internalize collaborative reasoning patterns and show improved TTS performance in MAS. Furthermore, we propose an adaptive scaling strategy incorporating a ``CEO'' agent to dynamically guide the reasoning process and optimize collaboration depth. Extensive experiments within the AgentVerse framework demonstrate that our fine-tuned models, Qwen2.5-32B-MAS and Phi4-14B-MAS, significantly outperform their base counterparts. Codes are available at https://github.com/jincan333/MAS-TTS.

cs.AI↗

arXiv2026

AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning

Test-time scaling strategies for Large Language Models predominantly rely on either reinforcement learning with sparse outcome rewards or search-based methods guided by static Process Reward Models. However, outcome-based RL often suffers from training instability and sample inefficiency, while static PRMs require expensive step-wise supervision and are susceptible to reward hacking due to distributional shifts. In this paper, we introduce AIRL-S, a unified framework that integrates Adversarial Inverse Reinforcement Learning with Group Relative Policy Optimization. By inferring a dense, step-wise reward model directly from reference trajectories, AIRL-S eliminates the dependency on labeled process data and uses the same learned PRM as both a training signal and a verifier for search-based TTS. Extensive evaluations across eight benchmarks in mathematics, science, and code generation demonstrate that our policy model improves average performance by 9\% over the base model, matching GPT-4o. We further analyze how the AIRL and GRPO objectives complement each other and how the learned PRM transfers across generators and search algorithms, establishing a robust and cost-effective methodology for scaling test-time computation in complex reasoning tasks.

cs.LG↗

arXiv2026

Uniqueness for the Degenerate Monge-Ampère Equation on Arbitrary Bounded Convex Domains

Let $n\ge2$ and let $Ω\subset\mathbb R^n$ be an arbitrary bounded open convex set. The author prove that, for $p>n$, the Dirichlet problem \[ \det D^2u=(-u)^p\quad\text{in }Ω, \qquad u=0\quad\text{on }\partialΩ, \qquad u>0\quad\text{in }Ω\] has at most one convex Alexandrov solution. The proof is based on the affine behavior of the Monge--Ampère energy and on a power-concavity property of the $L^{p+1}$ mass along the Legendre path connecting two solutions. At the homogeneous exponent $p=n$, the same argument shows that any two nonzero solutions with the same coefficient are positive multiples of one another.

math.AP↗

arXiv2026

ES-VP : Energy-Shaped Dynamic Visual Prompting for Efficient Model Adaptation

Visual prompting (VP) has emerged as a parameter-efficient method for adapting pre-trained models to downstream tasks. However, existing approaches encounter a trade-off between flexibility and efficiency. Some methods apply a fixed prompt to all images, ignoring individual image characteristics, while others introduce auxiliary networks to generate diverse prompts. Although the latter can improve performance, it also significantly increases parameter usage and the potential for overfitting to specific datasets. Furthermore, the auxiliary networks, combined with inherent biases in pre-trained models, limit scalability and generalization. In this paper, we propose Energy-Shaped Visual Prompting (ES-VP), a novel approach that generates image-specific prompts using low-rank initialization and energy-guided dynamic adaptation, achieving superior performance with fewer parameters compared to single-prompt methods. ES-VP directly utilizes the pre-trained model for adaptive prompt generation, ensuring both parameter efficiency and improved generalization. Extensive experiments conducted on five architectures across fifteen datasets demonstrate that ES-VP consistently outperforms current state-of-the-art (SOTA) single and diverse VP methods. For instance, using the CLIP architecture across four datasets, ES-VP outperforms the SOTA method DAM-VP by an average of 2.6\% in accuracy while utilizing 590$\times$ fewer VP parameters, thereby establishing a new benchmark for efficient and generalizable model adaptation.

cs.CV↗

arXiv2026

Pion structure in Holographic QCD

We employ a holographic model with a modified background that incorporates effective descriptions of key QCD features, including linear confinement and gluon condensation, to study the pion's internal structure, encompassing its mass spectrum as well as electromagnetic and gravitational form factors. This model is capable of simultaneously describing these diverse observables and reaches reasonable agreement with both experimental measurements and lattice QCD results. Our findings indicate that the model captures essential aspects of the pion. The description of multiple structure observables supports its potential as a useful tool for further investigations of pion properties.

hep-ph↗

arXiv2026

A Forward Simulation-Based Hierarchy of Linearizable Concurrent Objects

In this paper, we systematically investigate the connection between linearizable objects and forward simulation. We prove that the sets of linearizable objects satisfying wait-freedom (resp., lock-freedom or obstruction-freedom) form a bounded join-semilattice under the forward simulation relation, and that the sets of linearizable objects without liveness constraints form a bounded lattice under the same relation. Thus, forward simulation is not only a proof technique for linearizability but also induces an algebraic hierarchy of linearizable objects. As part of our lattice result, we propose an equivalent characterization of linearizability by reducing checking linearizability w.r.t. sequential specification $Spec$ into checking forward simulation w.r.t. an object $\mathcal{U}_{Spec}$.

cs.DC↗

arXiv2026

OODBench: Out-of-Distribution Benchmark for Large Vision-Language Models

Existing Visual-Language Models (VLMs) have achieved significant progress by being trained on massive-scale datasets, typically under the assumption that data are independent and identically distributed (IID). However, in real-world scenarios, it is often impractical to expect that all data processed by an AI system satisfy this assumption. Furthermore, failure to appropriately handle out-of-distribution (OOD) objects may introduce safety risks in real-world applications (e.g., autonomous driving or medical assistance). Unfortunately, current research has not yet provided valid benchmarks that can comprehensively assess the performance of VLMs in response to OOD data. Therefore, we propose OODBench, a predominantly automated method with minimal human verification, for constructing new benchmarks and evaluating the ability of VLMs to process OOD data. OODBench contains 40K instance-level OOD instance-category pairs, and we show that current VLMs still exhibit notable performance degradation on OODBench, even when the underlying image categories are common. In addition, we propose a reliable automated assessment metric that employs a Basic-to-Advanced Progression of prompted questions to assess the impact of OOD data on questions of varying difficulty more fully. Lastly, we summarize substantial findings and insights to facilitate future research in the acquisition and evaluation of OOD data.

cs.CV↗

arXiv2026

When Vision Becomes Text: Visual Token Pruning via Cross-Modal Residual Guidance in VLMs

Abundant visual information strengthens vision-language model (VLM) perception, yet massive visual tokens raise inference costs. Existing visual token pruning methods rely on similarity-based guidance, which exploits pairwise text-vision and vision-vision token correlations for compression. However, such methods only capture local layer-level signals and overlook the whole inference process in VLM. In this paper, we revisit VLM inference and present a new efficient guidance scheme that complements similarity-based guidance. In particular, we identify a key observation: as LLM layers deepen, text tokens continuously aggregate visual information via self-attention and progressively absorb partial visual content into textual representations. To quantify this phenomenon, we propose Cross Modal Absorption (CMA) from a geometric representation perspective to measure how much visual information is absorbed by text, revealing that more visual tokens in deeper layers can be approximately explained by the text subspace. We accordingly propose Cross Modal Residual (CMR). It projects visual tokens onto the text subspace via Tikhonov regularized least squares and exploits reconstruction residuals to quantify visual information that cannot be explained by text. Finally, based on CMR, we present SIEVE, a training-free visual token compression method that combines CMR, text-attention relevance, and residual-space diversity to retain task-relevant and complementary tokens. Experiments on diverse VLM architectures verify the effectiveness of SIEVE. For instance, on LLaVA-NeXT-7B, SIEVE keeps only $11.1\%$ of visual tokens while preserving $97.5\%$ of the original average performance, achieving $3.62\times$ prefill speedup, $2.49\times$ end-to-end speedup, and a $6.02\times$ KV-cache reduction.

cs.CV↗