Searcharxiv⌕ Search

arXiv subjects

Jiaqi Hou

Publications and source records attributed to Jiaqi Hou.

10 recordsLinked to original sources
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

Higher rank spectral projector estimates and a Stein--Tomas restriction theorem

Let $X$ be a compact locally symmetric space of noncompact type. We prove sharp upper bounds for the $L^2\to L^p$ operator norms of spectral projectors for joint eigenfunctions of the full ring of invariant differential operators on $X$. This improves the corresponding result of Marshall, by removing a logarithmic power loss at the critical exponent. Using the same method, we also prove the endpoint case of Mockenhaupt's $(L^2,L^p)$ Fourier restriction theorem.

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

TexTailor: Texture-Preserving Video Virtual Try-On via Adaptive Garment Conditioning

Video virtual try-on has attracted increasing attention due to its broad potential in digital fashion and intelligent e-commerce. However, existing methods primarily focus on low-resolution settings and still face substantial challenges when extended to high-resolution scenarios. These limitations can be attributed to two main factors: (1) the insufficient utilization of rich garment reference information, and (2) the lack of explicit positional modeling between garment and video representations during cross-modal interaction, which weakens fine-grained local correspondence. To address these issues, we propose TexTailor, a high-fidelity video virtual try-on framework built upon a pretrained video Diffusion Transformer. Specifically, we introduce a timestep-adaptive modulation mechanism to dynamically adjust garment visual representations throughout denoising. We further develop a frame-aligned positional encoding strategy to strengthen garment-to-video correspondence, together with a multi-source injection design that reduces interference among heterogeneous conditions. Extensive experiments on multiple video virtual try-on benchmarks, including the high-resolution Eevee dataset, demonstrate that TexTailor achieves competitive performance in garment detail preservation, temporal consistency, and overall video quality.

cs.CV↗
arXiv2026

$L^q$-norm bounds for arithmetic eigenfunctions via microlocal Kakeya-Nikodym estimate

Let $X$ be a compact arithmetic congruence hyperbolic surface, and let $ψ$ be an $L^2$-normalized Hecke-Maass form on $X$ with sufficiently large spectral parameter $λ$. We give a new proof to obtain a power saving for the global $L^6$-norm $\|ψ\|_{L^6(X)}\lesssim_\varepsilonλ^{\frac{5}{36}+\varepsilon}$ over the local bound $\|ψ\|_{L^6(X)}\lesssimλ^{\frac{1}{6}}$ of Sogge. Our method uses a microlocal decomposition for $ψ$ and reduces the $L^6$-norm problem to microlocal Kakeya-Nikodym estimates for $ψ$, and we establish improved microlocal Kakeya-Nikodym estimates via arithmetic amplification developed by Iwaniec and Sarnak.

math.NT↗
arXiv2025

Restrictions of Maass forms on $\mathrm{SL}(2,\mathbb{C})$ to hyperbolic surfaces and geodesic tubes

Let $ψ$ be an $L^2$-normalized Hecke-Maass form with a large spectral parameter $λ>0$ on a compact arithmetic congruence hyperbolic 3-manifold $X=Γ\backslash\mathrm{SL}(2,\mathbb{C})/\mathrm{SU}(2)$, and let $Y$ be a totally geodesic surface in $X$ with bounded diameter. The local $L^2$-bound for the restriction of $ψ$ to $Y$ is $\|ψ|_Y\|_{L^2(Y)}\ll λ^{1/4}$ by Burq, Gérard, and Tzvetkov. We apply the method of arithmetic amplification developed by Iwaniec and Sarnak to obtain a power saving over the local bound. The new feature in the proof is that we establish two different estimates for the integrals of $ψ|_Y$ against geodesic beams over $Y$ via two amplification arguments. Combining these estimates, we can improve the local bound for generalized Fourier coefficients of $ψ|_Y$ against eigenfunctions on $Y$ with spectral parameters near $λ$. We also apply the amplification method to obtain a power saving over the trivial bound $O(1)$ for $L^2$-norms of $ψ$ restricted to $λ^{-1/2}$-neighborhoods of unit-length geodesic segments. Consequently, by applying a result of Blair and Sogge, we obtain power savings over the local $L^p$-bounds of $ψ$ by Sogge for $2<p<4$ from our improved bound for the Kakeya-Nikodym norm.

math.NT↗
arXiv2025

Kakeya-Nikodym norms of Maass forms on $\rm{U}(2,1)$

Let $ψ$ be a Hecke-Maass form with a large spectral parameter on a compact arithmetic complex hyperbolic surface. We apply the amplification method to obtain a power saving over the trivial bound for the Kakeya-Nikodym norm of $ψ$. As a consequence, we obtain power savings over the local bound of Sogge for $\|ψ\|_p$ when $2<p<10/3$.

math.NT↗
arXiv2025

Weighted geodesic restrictions of arithmetic eigenfunctions

Let $X$ be an arithmetic hyperbolic surface, $ψ$ a Hecke-Maass form, $\ell$ a geodesic segment on $X$, and $μ$ a Borel measure supported on $\ell$ with dimension greater than 1/2. We obtain a power saving over the local bound of Eswarathasan and Pramanik for the $L^2$ norm of $ψ$ with respect to $μ$, which is a weighted generalization of Marshall's geodesic restriction bound and is proved by applying the method of arithmetic amplification. On a general 2-dimensional Riemannian manifold, we also obtain a Kakeya-Nikodym bound for the $L^2$ norm of any Laplace-Beltrami eigenfunction with respect to a Borel measure supported on a geodesic segment with dimension greater than 1/2.

math.NT↗
arXiv2024

A nonabelian circle method

We count integral quaternion zeros of $γ_1^2 \pm \dots \pm γ_n^2$, giving an asymptotic when $n\ge 9$, and a likely near-optimal bound when $n=8$. To do so, we introduce a new, nonabelian delta symbol method, which is of independent interest. Our asymptotic at height $X$ takes the form $cX^{4n-8} + O(X^{3n+\varepsilon})$ for suitable $c \in \mathbb{C}$ and any $\varepsilon>0.$ We construct special subvarieties implying that, in general, $3n+\varepsilon$ can be at best improved to $3n-2.$

math.NT↗