arXiv subjects

Jun Zhou

Publications and source records attributed to Jun Zhou.

At least 19 recordsLinked to original sources

arXiv2026

Latent Spectral Prompting for Multimodal Hyperspectral Object Tracking

Hyperspectral videos provide rich spectral signatures for robust tracking under challenging conditions such as camouflage and low illumination. However, limited hyperspectral annotations and the large RGB-hyperspectral domain gap hinder the direct use of modern temporally pretrained RGB trackers. We propose SProTrack, a multimodal framework that transfers temporal knowledge from pretrained RGB trackers to the hyperspectral domain through latent spectral prompting. The key idea is to decouple temporal modeling from spectral adaptation, allowing temporal dynamics to be inherited from RGB foundation models while hyperspectral information is learned from unlabeled data in a self-supervised manner. Unlike band-selection methods that discard most spectral information or prompt-based approaches that rely on local-search crops, SProTrack learns a global, scene-level spectral representation from the full hyperspectral frame using an autoencoder. This latent representation is used to guide spectral prompt extraction while a pretrained RGB tracker preserves strong temporal modeling on RGB-compatible inputs. By encoding spectral-spatial structure into compact latent prompts, SProTrack bridges the RGB-hyperspectral modality gap without extensive retraining or dense supervision. Experiments on HOT2020 and HOT2023 demonstrate that SProTrack achieves the best overall performance, with strong results across all spectral domains.

cs.CV↗

arXiv2026

Efficient Hybrid Distillation: Synergizing Score and Adversarial Objectives for One-Step Diffusion

Numerous distillation methods have been developed to accelerate diffusion models. Recent research indicates that hybrid strategies---blending different paradigms---yield superior results compared to single-method approaches. However, current works predominantly rely on a simple combination of objectives, leaving the question of how to effectively synergize them largely underexplored. In this paper, we specifically investigate the synergy between Score Distillation and Adversarial Distillation. We reveal that their operating zones are distinct and complementary: Score Distillation is effective in high-noise regimes for capturing the global distribution yet lacks fine-grained supervision in low-noise regions. Conversely, Adversarial Distillation is effective in low-noise regions but is prone to instability and suffers from discriminator failure when applied to high-noise intervals. Leveraging this insight, we propose Efficient Adversarial Score Distillation (EASD), a framework that synergizes these objectives via a regime-adaptive sampling strategy, directing each objective to focus on its effective interval. We further apply this paradigm to DiT-based Flow Matching models. Empirically, our method achieves a promising one-step FID of 1.39 on ImageNet 256$\times$256 using the SiT-XL/2+REPA model. This work provides a concrete guideline for maximizing the potential of hybrid distillation. Our code and model are available at https://github.com/pf0607/EASD .

cs.CV↗

arXiv2026

Exponential stability of the reduced Bresse--Gurtin--Pipkin system with one temperature

We consider the reduced Bresse--Gurtin--Pipkin system with one temperature, obtained from the full Bresse--Gurtin--Pipkin thermoelastic beam system by neglecting one of the two temperature variables. This system was left essentially open in a recent work of Dell'Oro, where only a conjecture on its exponential stability condition was formulated. By combining the spectral analysis recently developed by Dell'Oro, Pata and Vasylyeva for the abstract Gurtin--Pipkin heat equation with the resolvent estimate framework of Dell'Oro for Bresse systems, we prove that the associated contraction semigroup is exponentially stable if and only if two explicit stability numbers, depending on the structural constants of the beam, both vanish. The proof is based on a spectral characterization of the infinitesimal generator in a half-plane containing the imaginary axis, an analysis of its spectrum on the imaginary axis, sharp upper bounds for the resolvent, and the Gearhart--Prüss theorem. In particular, we confirm the conjecture of Dell'Oro and provide a rigorous spectral analysis of the reduced system. The same method yields the exponential stability characterization of the corresponding reduced Bresse--Maxwell--Cattaneo system, and can be adapted to other reduced Bresse-type models.

math.AP↗

arXiv2026

Continuity of Global Attractors under a Mass-Preserving Singular Memory Limit, with an Application to a Coupled Suspension Bridge System

We study how the global attractor of a dissipative evolution equation with memory depends on the memory kernel when the kernel concentrates, while preserving its total mass, towards a Dirac mass at the origin. The mass-preserving rescaling is the structural feature that prevents the effective elastic coefficients from degenerating in the limit, so that the memory-dependent dynamics reduces formally to a memoryless damped system with unchanged elastic coefficients. For a coupled nonlinear suspension bridge system with past history, taken as a concrete model, we establish uniform-in-parameter dissipativity and uniform-in-parameter exponential stability of the associated linear semigroup. These uniform estimates yield uniform quasi-stability, uniform exponential attraction, finite-time convergence of the semigroups, and uniform Hölder continuity with respect to the perturbation parameter. We deduce upper semicontinuity of the global attractors, and, under an additional structural stability hypothesis on the limit attractor, lower semicontinuity as well. The analysis separates the general continuity mechanism, which applies to a broad class of memory kernels and dissipative systems, from the model-specific verification, which exploits the coupled bridge structure.

math.AP↗

arXiv2026

SWE-Trail: History-Guided Trajectory Collection for Coding Agent Training

Coding agents solve repository issues through long sequences of exploration, editing, and validation. These trajectories are increasingly used as post-training data, so how they are collected shapes what the trained model learns. A common approach is to sample multiple trajectories independently for each issue and then select or curate them. However, as more trajectories are sampled independently, later trajectories become increasingly redundant, even though earlier trajectories could guide where subsequent collection effort should be spent. A natural way to use this history is to expose it to the coding agent as advice. But training on such trajectories can make the model rely on advice that is unavailable at test time, creating a train-test mismatch. Instead, we argue that history should guide where collection effort is spent, while the coding agent decides what to generate. We introduce SWE-Trail, a trajectory collector that uses earlier trajectories to guide subsequent collection while keeping that history outside the coding agent's context. During a new rollout, SWE-Trail uses history to identify a state where additional collection effort may be useful. There, the coding agent proposes candidate next actions from its ordinary context, and SWE-Trail briefly executes them from the same checkpoint to decide which route continues. We evaluate collection approaches by training student models on the trajectories they produce. For a 124B student initialized from Ling-3.0-flash-base, SWE-Trail improves mean resolution over independent sampling by 2.33 and 1.60 percentage points on SWE-bench Multilingual and SWE-bench Pro, respectively. Without history, the same pipeline produces more resolved trajectories than independent sampling, but improves by only 0.11 points on Multilingual. These results show the value of using history to guide collection without exposing it to the coding agent.

cs.AI↗

arXiv2026

Fight World Model: Customizable Combat through Visual Interaction Proxies

A playable, programmable game world must maintain persistent state, enforce executable rules, and support extensible gameplay mechanics. Existing video world models can generate plausible combat, yet struggle to keep hit reactions and health updates consistent with game rules. We introduce Fight World Model (FWM), which separates rule-governed combat execution from video rendering. An executable engine maintains state and resolves interactions. A deterministic compiler encodes these outcomes and attempted actions in a spatially and temporally aligned interaction proxy. The proxy aligns action and skill windows, brief hit reactions, and persistent health by fighter identity and timing. These cues can be controlled separately to render interactions under modified rules. Identity-bound references preserve character appearance across side switches. We train FWM on FightWorld Corpus, spanning 52 skill effects, 187 character-appearance variants, and 15 arenas. Without retraining the renderer, FWM supports damage scaling, limited-hit immunity, lifesteal, and multi-stage skill mechanisms beyond the native game. It also allows character and arena customization. Experiments demonstrate the effectiveness of FWM, which outperforms existing game world models in action fidelity, hit reactions, health-state consistency, and character appearance preservation. Our project is available at https://fight-world-model.github.io.

cs.CV↗

arXiv2026

CausalMix: Data Mixture as Causal Inference for Language Model Training

In Large Language Model (LLM) training, data mixing plays a pivotal role in determining model performance. Recent methods optimize mixture weights via proxy models, but they rely on the assumption of static data distributions. As a result, when the underlying data pool shifts, these methods require costly retraining from scratch. This limitation restricts their ability to scale seamlessly from small settings to larger data pools and model sizes. In this paper, we propose CausalMix to address this limitation by casting data mixture optimization as a causal inference problem. We formulate the statistical features of the data pool as covariates and the domain mixture as the treatment. After fitting a causal model on 512 runs of Qwen2.5-0.5B to estimate the Conditional Average Treatment Effect (CATE), we extrapolate the optimal mixture for an 800K data pool and apply it to train a 7B model. Furthermore, we successfully generalize the framework to long chain-of-thought data on Qwen3-4B-Base. By leveraging causal modeling to isolate confounding biases, CausalMix infers state-dependent data mixtures offline, with the selected mixture fixed during each training run. Extensive experiments show that the mixture guided by CausalMix achieves competitive performance across downstream tasks compared with other baselines. The CATE Interpreter visualizes the learned mixing strategy. Overall, CausalMix offers a causal and interpretable framework for optimizing LLM data mixtures.

cs.LG↗

arXiv2026

Foundation Models in Medical Image Analysis: A Systematic Review and Meta-Analysis

Recent advancements in foundation models (FMs) have catalyzed a paradigm shift in medical image analysis. Unlike traditional task-specific artificial intelligence (AI) models, FMs leverage large-scale datasets to learn generalized representations that can be adapted to downstream clinical applications. Despite the rapid proliferation of FM research in medical imaging, there is a lack of unified synthesis that systematically maps the evolution of architectures, training paradigms, and clinical applications across modalities. To address this gap, this review provides a comprehensive and structured synthesis of FMs in medical image analysis by systematically organizing studies into two primary categories: vision-only foundation models (VFMs) and vision-language foundation models (VLFMs), based on their architectural foundations, training strategies, and downstream clinical tasks. A quantitative analysis was conducted on both VFMs and VLFMs to characterize temporal trends in dataset utilization and application domains, along with pooled performance and subgroup analyses. We also critically discuss persistent challenges, including cross-domain generalization, computational scalability, FM evaluation, fairness, and deployment. Finally, we identify key future research directions aimed at enhancing the robustness, interpretability, and clinical integration of FMs, thereby accelerating their translation into real-world medical practice.

cs.CV↗

arXiv2026

AV-FABLE: Audio-Visual Manipulation Forensics in Long-Form Videos

Localized AI manipulations in long-form videos often affect brief events within otherwise authentic content, making video-level classification insufficient for forensic analysis. Existing audio-visual deepfake benchmarks largely focus on short talking-head clips, while long-form temporal forensics remains predominantly visual. We introduce AV-FABLE, a dual-timeline benchmark for sparse manipulations in open-scene long-form audio-visual videos. AV-FABLE contains 16,650 pristine--manipulated twin pairs (998 hours) created with 27 editing configurations, with separate audio and visual timelines that enable modality-specific temporal localization and expose errors hidden by joint-timeline evaluation. On AV-FABLE, zero-shot prompting of audio-visual large language models (AV-LLMs) yields near-chance detection and poor temporal localization. In contrast, supervised readouts of their frozen perception features recover strong forensic signals, suggesting manipulation evidence is encoded in these representations but not effectively exposed through language outputs. Motivated by this observation, we propose NFSR(Native Forensics via Separated Readouts), which reads frozen AV-LLM features with separate audio and visual temporal decoders. NFSR achieves the best audio-visual localization performance in-domain and across all three out-of-distribution settings, including transfer to held-out videos from an external long-video forensic benchmark. AV-FABLE, code, and model weights will be publicly released.

cs.CV↗

arXiv2026

OACap: Caption-Guided Active Observation for Detailed Video Captioning

Unlike question answering, where a query specifies a perceptual target, open-ended detailed captioning must determine what evidence to acquire and how to integrate it into a coherent global description. We propose Observation-Adaptive Video Captioner (OACap), a caption-centered agent that turns unsupported or missing information into targeted observation requests, uses spatiotemporal tools to acquire evidence, and iteratively verifies and revises the caption. OACap captions consistently improve performance across general video-understanding benchmarks using a fixed text-only reader. To better diagnose the information they preserve, we introduce Factual-Semantic Captioning Bench (FSCap-Bench). Its Core-Support Factuality Score (CSFS) combines claim precision with importance-aware fact coverage and responds as expected to four controlled perturbation types on 93.0-98.3% of valid caption pairs. Its Cross-Modal Semantic Round-Trip Score (CSRS) measures semantic preservation through a video-caption-video-caption round trip by matching atomic facts from source and recovered captions. Together, they connect downstream textual-memory utility with factual and semantic preservation.

cs.CV↗

arXiv2026

Tempo-Zero: Recursive Self-Play in Executable Worlds for Video Reasoning

Video reasoning requires understanding how events unfold over time, yet videos with accurate temporal annotations are expensive to collect. Self-play offers an alternative by letting a model generate and solve its own tasks without manual annotation. However, existing video self-play methods train on fixed sets of recorded videos, so they cannot alter the underlying event dynamics and rely on noisy pseudo-labels derived from the model's own outputs, such as majority voting, that risk reinforcing shared model errors. To overcome these limitations, we introduce Tempo-Zero, a self-improvement framework that closes a loop of executable environments, execution-verified self-play, and environment evolution, where executing a program renders a video with the ground-truth option and interval of every question. A single policy plays both Questioner and Solver, proposing questions and answering them with timestamped evidence lines, and is trained on execution-grounded rewards instead of model consensus. The Solver's performance steers a curriculum over tasks and difficulty levels, and the policy selects rule changes that enrich the event dynamics of later rounds. Despite being trained solely on simple programs, without supervised fine-tuning or external video data which prior methods typically require, Tempo-Zero outperforms the base models and recent self-play baselines on average across twelve real-world video benchmarks at different model scales, with particularly strong gains on temporal grounding tasks. Project page: https://alchemist-jo.github.io/tempo-zero/

cs.CV↗

arXiv2026

Well-posedness and Stability of Thermoelastic Suspension Bridges with Cattaneo Heat Conduction: The Role of Viscoelastic Memory

Thermoelastic suspension bridges operate under strong thermo-mechanical coupling, and the classical Fourier law fails to capture the finite speed of thermal propagation observed in low-conductivity materials such as concrete. This paper investigates a suspension bridge model in which the heat flux is governed by the Cattaneo law with relaxation time $τ$, and in which the bridge deck may or may not possess viscoelastic memory. Using operator semigroup theory, we establish well-posedness and derive sharp decay rates for both regimes. For the memoryless system ($δ=0$), we show that the semigroup is polynomially stable with rate $t^{-1/2}$ when the structural parameters satisfy $\frac{ρ_1}κ = \frac{ρ_2}{b}$, and that the rate deteriorates to $t^{-1/4}$ otherwise; moreover, exponential stability fails whenever $χ_0 \neq 0$, or $χ_0 = 0$ and $χ_1 = 0$, where $χ_0$ and $χ_1$ are parameter-dependent coefficients. In sharp contrast, when viscoelastic memory is present ($δ=1$), the semigroup becomes exponentially stable, regardless of the structural parameter ratio. This reveals a regularizing role of viscoelastic memory: it converts polynomial decay into exponential decay and removes the parameter-ratio obstruction. The results provide a theoretical basis for exploiting inherent viscoelasticity in long-span bridges and for optimizing deck parameters in memoryless structures.

math.AP↗

arXiv2026

Material and Defocus Cues for Occluded Object Localization in Single Hyperspectral Images: An Exploratory Study

Partial occlusion makes object detection difficult when visible regions have weak texture or similar appearance. Hyperspectral imaging provides material-dependent spectra and wavelength-dependent optical blur, which offer complementary evidence for separating objects. We investigate a pipeline that combines nonnegative-matrix-factorization-based material grouping, divergence-based optical cues, and graph Laplacian propagation. The propagated scores support region extraction and bounding-box localization without requiring a separate depth sensor at inference. A collection of 30 indoor and six outdoor hyperspectral scenes acquired using two camera systems provides qualitative examples spanning different materials, depth arrangements, and partial occlusions. The examples illustrate how material information can support separation when optical cues are weak, and how optical cues can distinguish some surfaces with similar material properties. Objects sharing both material and depth remain challenging. This exploratory study establishes the sensing rationale and mathematical formulation, while identifying the quantitative evaluation and implementation details needed to assess detection performance systematically.

cs.CV↗

arXiv2026

Global Well-Posedness and Energy Decay for an Abstract Viscoelastic Wave Equation with Nonlocal or Friction Damping

This paper investigates an abstract viscoelastic wave equation in a separable Hilbert space, in which the damping term depends on a power of the velocity norm. Depending on whether the power exponent is zero or nonzero, this damping corresponds to friction damping or nonlocal damping, respectively; the equation also contains a viscoelastic damping represented by an integral term with a relaxation function. We establish global well-posedness for initial data in $\mathcal{H}_1$ by combining monotone operator theory with the Faedo--Galerkin method, and prove uniqueness for all exponents greater than $-1$, with the singular range between $-1$ and $0$ handled through the monotonicity rather than the Lipschitz continuity of the nonlocal damping operator. We then study the long-time behavior of the energy via integral inequalities, multiplier techniques, and Nakao's method, and obtain explicit decay estimates showing that the energy decays exponentially in the friction case and polynomially in the nonlocal case. To the best of our knowledge, this is the first work to consider the combined effects of viscoelastic damping and nonlocal damping in such an abstract framework, and it is also the first to analyze nonlocal damping with negative exponents.

math.AP↗

arXiv2026

Exponential Stability of the Linear Semigroup and the Global Attractor for a Coupled Suspension Bridge System with Past History

This paper is concerned with the long-time dynamics of a coupled nonlinear suspension bridge system with past history, in which the deck and the main cable interact through the nonlinear restoring force $k[u-v]^{+}$. Building on the well-posedness and the solution semigroup established in \cite{Mukiawa2025}, we first show the exponential stability of the associated linear semigroup by means of the Gearhart--Prüss theorem together with a detailed resolvent estimate. We then combine this exponential stability with the local Lipschitz property of the nonlinearity, the fact that the nonlinearity acts only on the elastic component, the smoothing of the elastic component, and the exponential decay of the memory kernels to conclude that the solution semigroup is quasi-stable in the sense of Chueshov--Lasiecka, and therefore asymptotically smooth. Together with the dissipativity recalled from \cite{Mukiawa2025}, this leads to a compact global attractor in the natural energy space. In particular, our approach may be viewed as closing a gap in the proof of asymptotic smoothness in \cite{Mukiawa2025}, where a non-decaying initial-difference term was absorbed into an arbitrarily small constant without justification.

math.AP↗

arXiv2026

ContourVLA: A Closed-Loop Perception-Action Contour Policy for Generalized Referring Expression Segmentation

Generalized referring expression segmentation (GRES) requires dynamically balancing high-level semantics for identifying a variable number of language-specified referents with fine-grained visual evidence for precise boundary delineation. This requirement challenges existing cascaded vision-language architectures, which typically rely on static feature interfaces and single-pass mask prediction, limiting adaptive perception and geometric correction. We introduce ContourVLA, a vision-language-action policy that recasts GRES as a closed-loop visuomotor process, in which editable contours serve as explicit policy states that condition multimodal perception and are updated by geometric action chunks. Evolution-Aware Semantic Scheduling (EASS) couples contour-guided bidirectional boundary sampling with state-conditioned routing of multilevel multimodal features, adapting perception to each contour state. Following supervised initialization, Dustbin-Augmented Entropic Credit Transport GRPO (DECT-GRPO) jointly optimizes discrete grounding and continuous contour actions with instance-level credits. Its rollout rewards and credits are derived from soft prediction-target correspondences that account for false positives and missed targets. ContourVLA improves gIoU over the strongest evaluated baselines by 8.7, 2.8, and 2.7 points on gRefCOCO val, testA, and testB, respectively, and achieves the highest mIoU across all eight RefCOCO, RefCOCO+, and RefCOCOg splits.

cs.CV↗

arXiv2026

Aligning LLMs with Biomedical Knowledge using Balanced Fine-Tuning

Engineering LLMs to accelerate life sciences research requires a robust alignment with biomedical knowledge. We observe that biomedical text exhibits a fundamentally different uncertainty structure from general text: dense low-confidence runs encode epistemic knowledge gaps (dense causal chains, rare entities) rather than the sparse aleatoric stylistic variation typical of general text. Based on this discovery, we propose Balanced Fine-Tuning (BFT), a dual-scale post-training method that combines group-normalized token reweighting with sequence-level reallocation toward knowledge-dense samples exhibiting dense epistemic uncertainty. Across medical evaluation, biological reasoning, sparse-reward RL, and biological representation tasks, BFT provides more consistent gains than SFT and DFT under a shared training setup. When replacing the default closed-source backbones in GeneAgent (GPT-4o) and VCWorld (Gemini-2.5-Flash), the BFT-aligned 70B model delivers stronger performance across biological process reasoning and chemical perturbation prediction. Critically, all BFT variants further improve after subsequent GRPO with sparse rewards, while SFT and DFT degrade, suggesting that epistemic-aware post-training provides a more robust policy initialization. Beyond text generation, BFT-aligned LLMs produce more accurate and professional biomedical profile texts; after encoding these profiles with a text embedding model, the resulting representations support gene-level, cell-level, and perturbation-response tasks, suggesting that BFT-enhanced generation can facilitate biological representation and, in turn, broader biomedical downstream tasks.

cs.LG↗

arXiv2026

RADC: Risk-Aware Dual Caching for Vision-Language Test-Time Adaptation

Cache-based test-time adaptation (TTA) for vision-language models is often hindered by background bias in global representations and unreliable entropy-based cache admission under representation variations. To address these limitations, we propose RADC, which enhances prototype learning through reliable dual caching. RADC introduces a Semantic Foreground Cache that aggregates category-consistent spatial evidence from CLIP representations, yielding foreground prototypes that complement the global cache while mitigating background interference. To reliably manage both caches, Gaussian Risk Admission models multi-view representations as diagonal Gaussian distributions and jointly considers class separation and feature uncertainty to prioritize reliable cache candidates. RADC integrates zero-shot logits with complementary global- and foreground-cache predictions for robust inference. Extensive experiments on cross-domain and out-of-distribution benchmarks demonstrate consistent state-of-the-art performance.

cs.CV↗