arXiv subjects

Xingyu Zhu

Publications and source records attributed to Xingyu Zhu.

At least 19 recordsLinked to original sources

arXiv2026

NOUS: Video-Driven 3D Human Reaction Generation via Observation-Reaction Mutual Steering

Video-driven 3D human reaction generation aims to synthesize 3D human motion in response to the action observed in a video, playing an important role in interactive multimedia systems and embodied agents. Yet reaction motions generated by current methods often fail to match what the observed video calls for. We observe that one factor behind this failure is relational distortion in the correspondence between visual observations and reactions: videos lying close in the visual space may correspond to entirely different motions in the reaction space, which misleads the model into generating reactions inconsistent with the conditioning video. This motivates us to propose a new observatioN-reactiOn mUtual Steering (\texttt{NOUS}) framework that enables mutual steering between the video and motion modalities. It first performs Motion Feedback Steering (MFS), equipping the frozen pretrained video encoder with a lightweight rectification modulator and training the modulator with a relational margin loss that pulls each video embedding toward the motion prototype of its own category and away from those of other categories. In this way, the misaligned correspondence between visual observations and reactions can be calibrated. \texttt{NOUS} then applies Observation-Guided Refinement (OGR), which in turn exploits the rectified observations to further refine the generated reactions and enhance their quality. The results on the ViMo dataset demonstrate that \texttt{NOUS} improves the quality of reaction motion while incurring negligible computational overhead at inference. Also, \texttt{NOUS} yields consistent gains across four pretrained video encoders, showing its good compatibility.

cs.CV↗

arXiv2026

Positive Scalar Curvature and Volume Growth

For a complete Riemannian manifold with nonnegative Ricci curvature, we prove two sharp volume growth order estimates, thereby resolve a conjecture of Gromov in 1986. There first is that a uniform deficit in the volume of unit balls, an analog of positive macroscopic scalar curvature, forces codimension one volume growth, and the second one is that a uniformly positive scalar curvature lower bound forces codimension two growth known as the codimension two volume growth conjecture.

math.DG↗

arXiv2026

PEA-DPO: Perception-Enhanced Alignment Direct Preference Optimization for MLLMs Alignment

Direct Preference Optimization (DPO) has emerged as an effective approach for aligning large language models (LLMs) with human preferences. However, its adaptation to multimodal settings remains unexplored. Through representational analysis, we identify a key limitation in multimodal preference optimization, which we term visual insensitivity: models often fail to distinguish between images and those with critical visual context removed. Our theoretical analysis further uncovers two manifestations of this problem, namely Across-Image Insensitivity and Within-Image Insensitivity. To address these challenges, we propose Perception-Enhanced Alignment DPO (PEA-DPO), a framework for multimodal LLMs alignment, which explicitly leverages visual preference signals to overcome visual insensitivity. We further provide a theoretical analysis demonstrating that PEA-DPO provably mitigates both failure modes. Empirical results demonstrate that PEA-DPO enhances sensitivity to visual context while preserving the language modeling capacity of the base model. Evaluations across three hallucination benchmarks using MLLMs of varying scales show that PEA-DPO effectively mitigates visual insensitivity, achieves stronger multimodal alignment, and substantially reduces hallucinations.

cs.CV↗

arXiv2026

FlatLab: A Unified Methodology Framework and Simulation-Based Benchmark for Robotic Manipulation of Flat Objects

Robotic manipulation of flat objects is challenging due to the ungraspable configurations and strong variations in object geometry and material. Existing methods rely on heuristic pre-manipulation and are often evaluated in closed settings with limited generalization. We propose a unified framework that decouples the manipulation into a strategy generator and an action execution module. The strategy generator predicts appropriate manipulation strategies from object point clouds by learning strategy-centric, object-invariant representations via simulated data transformation and contrastive learning. Conditioned on the predicted strategy, the execution module decomposes long-horizon manipulation into reusable action primitives and dynamically composes them to generate stable trajectories. To enable systematic evaluation, we introduce FlatLab, a comprehensive simulation benchmark for robotic flat object manipulation. FlatLab provides high-fidelity physical simulation of diverse rigid and deformable flat objects, automated multi-modal data collection, and standardized task definitions and evaluation protocols. Experiments conducted in FlatLab demonstrate that our approach generalizes effectively to unseen objects and categories, outperforming existing baselines. The project page and the code are provided at https://flatlab-web.github.io/.

cs.RO↗

arXiv2026

Search for quasar pairs with Gaia astrometric data IV. Confirmation of 17 dual quasars and 143 projected quasars

Dual quasars separated at the kiloparsec scale are widely regarded as precursors to binary supermassive black holes and offer a key insight into the dynamical evolution of galaxy mergers. Our series of studies focuses on searching for dual quasars by using a selection strategy of zero proper motion and zero parallax to isolate candidate quasars near known ones and by follow-up spectroscopy of the candidates. This paper, the fourth in the series, reports the spectroscopic confirmations of our quasar pair candidates based on the spectroscopic data of the SDSS and DESI DR1. We newly identified 17 dual quasars and 143 projected quasars. The redshifts of the 17 dual quasars range from 0.573 to 2.758, with a median of 1.512. One notable system, J0023+0417, exhibits nearly identical spectral features in the two members and shows evidence of a potential foreground galaxy, making it a high-confidence strong gravitational lensing system. The redshifts of the 143 projected quasars are from 0.301 to 4.030, with a median of 1.596. Among them, four have projected distances below 30 kpc, offering valuable opportunities to probe the circumgalactic medium (CGM) of the foreground host galaxy through absorption lines.

astro-ph.GA↗

arXiv2026

T^2MLR: Transformer with Temporal Middle-Layer Recurrence

Transformer reasoning is limited by autoregressive decoding, which repeat edly compresses rich hidden computation through token space and makes it difficult for intermediate reasoning states to persist across time. We in troduce Transformers with Temporal Middle-Layer Recurrence (T2MLR), a transformers-based latent reasoning architecture that fuses a cached middle layer representation from the previous token directly into an earlier layer of the current token position, enabling abstract intermediate computation to persist across decoding steps with little inference overhead. Across natural-language pretraining and multi-hop reasoning finetuning, T2MLR consistently outperforms data- and parameter-matched Transformer base lines. Moreover, applying recurrence to only a localized middle-layer block (as little as 20% of the network) often outperforms full-layer recurrence. Im portantly, T2MLR does not require pretraining from scratch: retrofitting the recurrent pathway into an existing pretrained 1.7B Transformer and briefly finetuning substantially improves math reasoning, lowering the barrier to practical adoption. These results suggest that effective latent reasoning in Transformers does not require looping over all layers as in previous works, but can instead emerge more strongly from targeted middle-layer recurrence.

cs.CL↗

arXiv2026

Robustifying Vision-Language Models via Test-Time Prompt Adaptation

Pre-trained Vision-Language Models (VLMs) such as CLIP achieve strong zero-shot generalization, but their performance degrades sharply under adversarial perturbations. Existing test-time adaptation methods typically rely on sample-level confidence heuristics, overlooking the intrinsic distributional structure of the data. This sample-centric approach limits robustness, as it fails to distinguish confident adversarial mispredictions from true semantic consistency. In this work, we observe that adversarial distortion is structurally brittle: while holistic representations are corrupted, semantic integrity is often preserved in the distribution of augmented views. Motivated by this insight, we propose RITA, a Robust test-tIme prompt-TAdaptation framework that shifts from sample-level estimates to distribution-level alignment. Specifically, RITA employs optimal transport to align the distribution of augmented visual features with textual prototypes, mitigating adversarial outliers and rectifying cross-modal semantic misalignment. Furthermore, we introduce a dynamic cache to progressively accumulate reliable cues from the test stream for online refinement. Extensive experiments demonstrate that RITA significantly improves adversarial robustness without compromising clean accuracy.

cs.CV↗

arXiv2026

Search for Quasar Pairs with ${\it Gaia}$ Astrometric Data. III. Discovery of 9 dual and projected quasars

We report the low-resolution long-slit spectroscopic observations and confirmations of 11 quasar pair candidates, which are selected from the MGQPC catalog presented in the first paper of our series work (hereafter, Paper-I) and the early version of this catalog. The spectroscopic follow-up was carried out with 5 spectrographs equipped on 3 telescopes, and the major discoveries include 6 dual quasars and 3 projected quasars. One of the dual quasars has a high redshift of $\sim$ 3.1. The LQ hypothesis of 3 dual quasars cannot be completely ruled out. We investigated the reason why previous spectroscopic surveys missed several new quasars. We discussed a projected quasar with a wide-separation lensing configuration, as well as two quasar-star projections that mimic the configuration of lensed quasars. The photometric redshifts for the 11 observed candidates were extracted from the second paper of our series work (hereafter, Paper-II) to illustrate their positive role in mitigating contamination from projected quasars and quasar-star projections. We also reviewed and discussed the confirmation strategies for dual and lensed quasar candidates, and outlined future confirmation strategies for them in the context of the era dominated by large-scale spectroscopic and imaging surveys.

astro-ph.GA↗

arXiv2026

Nonnegative Ricci curvature and virtual abelianness in dimensions less than 12

For any complete Riemannian manifold $M^n$ with nonnegative Ricci curvature and sublinear diameter growth, we establish a dimensional constraint $n\ge 4s(s-1)+k+1$ if the fundamental group $π_1(M)$ contains a torsion-free nilpotent subgroup of rank $k$ and step $s\ge 2$. As a consequence, if such a manifold $M$ has dimension $n<12$, then $π_1(M)$ is almost abelian. The proof is based on a dimensional estimate for $\mathrm{RCD}(0,N)$ spaces admitting $\mathbb{R}$-orbits of large Hausdorff dimension.

math.DG↗

arXiv2026

Dual-Route Top-K Retrieval with 1v1 VLM Reranking for the CoVR-R

We describe \emph{Dual-Route Top-K Retrieval with 1v1 VLM Reranking} for the CoVR-R challenge. The method treats composed video retrieval as two coupled problems: finding a sufficiently complete top-k candidate set, and then safely deciding whether any candidate should replace a strong current top-1. We first improve the reasoning/text seed with a VLM slot selector over existing candidates, without introducing DFN visual retrieval. We then add a visual route from contact-sheet embeddings using DFN-H/DFN-L. The routes are merged into a top-10 candidate set, after which a VLM final reranker performs conservative 1v1 comparisons between the current top-1 and each challenger. On the hidden test split, the final system reaches 95.28 R@1, 97.47 R@5, 98.48 R@10, and 99.66 R@50. The main lesson is that CoVR-R benefits more from recall-selection decoupling than from broad text reranking or direct multi-candidate VLM classification.

cs.CV↗

arXiv2026

Adaptive Dense Evidence Refinement for Video Relational Reasoning for VRR-QA Challenge

VRR-QA evaluates whether video-language systems can infer spatial, temporal, viewpoint, depth, and visibility relations that are not always resolved by a single frame. We present an inference-only system built around adaptive test-time computation. The system first answers each question with a direct video-language model pass, then uses multiple lightweight views to find unstable questions. Only these difficult questions are routed to a high-budget dense evidence module that constructs timestamped frame observations, relation-specific probes, candidate verification, and conservative temporal aggregation. This design separates two problems that are often confused in video question answering: finding plausible alternative answers and deciding when a current answer should actually be changed. On the test split, the final system obtains 90.07 average accuracy and 87.81 macro average accuracy. The report focuses on the final test system and the implementation settings required to reproduce the adaptive dense verifier.

cs.CV↗

arXiv2026

Temporal Evidence Routing with Structured Visual Evidence for TimeLogicQA

TimeLogicQA evaluates whether video question answering systems can reason over temporal relations such as event existence, ordering, persistence, boundary conditions, and overlap. We address this task with a visual evidence routing pipeline that separates perception from symbolic temporal reasoning. The system first parses each question into event targets, answer mode, candidate options, and temporal operators. It then routes videos according to duration and operator difficulty, using ordered full-frame evidence for short clips and event-focused candidate windows for long videos. A multimodal large language model produces structured visual evidence for the relevant events, while programmatic verifiers recover dense action intervals and a deterministic reducer applies operator-specific temporal rules to produce the final answer. Conservative fusion accepts an answer only when the visual evidence, temporal program, and confidence checks agree, reducing noisy answer flips. On the official test evaluation, our final system achieves an AvgAcc of 81.8.

cs.CV↗

arXiv2026

Nonnegative Ricci Curvature and Uniformly Convex Boundary Forces Compactness

We confirm a compactness conjecture of M. Li. If a complete Riemannian manifold has nonnegative Ricci curvature and uniformly convex boundary in the sense that the second fundamental form satisfies $h\ge1$. Then we prove it is compact, and consequently has finite fundamental group. The proof uses monotone quantities constructed via positive proper harmonic functions with Neumann condition.

math.DG↗

arXiv2026

TCP-MCP: Landscape-Guided Co-Evolution of Prompts and Communication Topologies for Multi-Agent Systems

Effective multi-agent systems cannot be designed by selecting prompts or communication graphs in isolation. Agent behavior depends on the information an agent receives, while the usefulness of a communication edge depends on how the receiving agent interprets and uses that information. We propose \textbf{TCP-MCP} (Topology-Coupled Prompting for Multi-Agent Collaborative Problem-Solving), a co-evolution framework that searches agent prompts and communication topologies as a unified genome. TCP-MCP uses an initialization-time landscape probe to calibrate early search behavior, and then relies on Pareto-front diagnostics to adapt exploration under three objectives: task performance, token cost, and structural complexity. Using the same DeepSeek-V3.2 backbone across all methods, TCP-MCP achieves 82.66\%, 89.96\%, and 96.61\% accuracy on MMLU-Pro, MMLU, and GSM8K, respectively. Across the three benchmarks, it consistently outperforms automated graph-generation baselines and achieves competitive accuracy relative to debate-style systems, while using up to 5.69$\times$ fewer tokens than those systems at the reported operating points. These results show that jointly evolving prompts and communication structure provides a practical route to cost-aware and task-adaptive multi-agent system design in controlled evaluations.

cs.AI↗

arXiv2026

Fibrations, the First Betti Number, and Almost Nonnegative Ricci Curvature

In this paper, we prove fibration theorems for manifolds with almost nonnegative Ricci curvature and certain extra regularity assumptions. We show that a closed $n$-manifold $M$ satisfying $\mathrm{diam}(M)^2\mathrm{sec}_M \geq -κ$ and $\mathrm{diam}(M)^2\mathrm{Ric}_M \geq -δ$, where $δ>0$ is sufficiently small depending only on $n$ and $κ$, fibers over a $b_1(M)$-torus. This removes the upper sectional curvature bound required in the earlier result of Yamaguchi \cite{Y88}. As a corollary, we obtain a refinement of Yamaguchi's smooth fibration theorem (\cite{Y91}), showing that the fiber itself (rather than a finite cover of it) fibers over a $b_1$-torus. Our results extend to manifolds satisfying a generalized Reifenberg condition introduced in \cite{HH24}, which encompasses both a lower bound on sectional curvature and the local rewinding Reifenberg condition. In the nonsmooth setting, a similar result also holds for a non-collapsed $\mathrm{RCD}(-ε(D,r,n),n)$ space whose diameter is bounded by $D$ and which satisfies the $(r,δ(n))$-local rewinding Reifenberg condition. The proofs rely on an equivariant regularity theorem for almost submetries under a lower Ricci curvature bound. In addition, we study the stability of rank of Abelian actions along equivariant Gromov-Hausdorff convergence in this paper.

math.DG↗

arXiv2026

Revisiting Robustness for LLM Safety Alignment via Selective Geometry Control

Safety alignment of large language models remains brittle under domain shift and noisy preference supervision. Most existing robust alignment methods focus on uncertainty in alignment data, while overlooking optimization-induced fragility in preference-based objectives. In this work, we revisit robustness for LLM safety alignment from an optimization geometry perspective, and argue that robustness failures cannot be addressed by data-centric methods alone. We propose \textit{ShaPO}, a geometry-aware preference optimization framework that enforces worst-case alignment objectives via selective geometry control over alignment-critical parameter subspace. By avoiding uniform geometry constraints, ShaPO mitigates the over-regularization that can harm robustness under distribution shift. We instantiate ShaPO at two levels: token-level ShaPO stabilizes likelihood-based surrogate optimization, while reward-level ShaPO enforces reward-consistent optimization under noisy supervision. Across diverse safety benchmarks and noisy preference settings, ShaPO consistently improves safety robustness over popular preference optimization methods. Moreover, ShaPO composes cleanly with data-robust objectives, yielding additional gains and empirically supporting the proposed optimization-geometry perspective. The code is available at https://github.com/liujilong0116/ShaPO.

cs.LG↗

arXiv2026

Asymptotic-Type Dimension Bounds through Combinatorial Approaches

We develop a probabilistic framework for large-scale dimension bounds in metric geometry, based on padded decompositions, randomized ball carving on net graphs, and the Lovász Local Lemma. For metric measure spaces with volume doubling constant $C_{\mathsf D}$, we prove the sharp bound $\mathrm{asdim}_{AN}(X)\le \mathrm{dim}_{AN}(X)\le \lfloor{\log_2 C_{\mathsf D}}\rfloor$. In particular, if $(M,g)$ is a complete Riemannian $n$-manifold with $\mathrm{Ric}_g\ge 0$, then $\mathrm{asdim}(M)\le n$, thereby settling a question of Papasoglu on manifolds with nonnegative Ricci curvature. We also show that if $(X,\mathsf{d},\mathfrak{m})$ is proper, volume noncollapsed, and has polynomial volume growth rate $ρ^V(X)$, then $\mathrm{asdim}(X)\le \lfloor{ρ^V(X)}\rfloor$. Moreover, the corresponding control function can be chosen to have polynomial growth. This extends Papasoglu's sharp asymptotic-dimension bound from graphs of polynomial growth to a metric-measure setting. As applications, we study equality in the polynomial-growth bound for universal covers of nilmanifolds, and under nonnegative Ricci curvature we relate the equality case in the volume-doubling bound to Gromov largeness, obtaining in particular a consequence for complete manifolds with positive scalar curvature.

math.MG↗

arXiv2026

Search for quasar pairs with Gaia astrometric data. II. Photometric redshift prediction with machine learning for the MGQPC catalogue

The identification of physically associated kiloparsec-scale quasar pairs is important for understanding galaxy evolution, the growth of supermassive black holes, and their co-evolution with host galaxies. However, their rarity and the high contamination from stellar superpositions and projected alignments require efficient pre-selection methods. We develop a machine-learning framework to produce photometric-redshift point estimates and redshift probability density functions for quasars, with the main goal of identifying high-probability quasar pair candidates in the MGQPC catalogue. We construct two large spectroscopically confirmed quasar samples with multi-wavelength photometry, based on SDSS and DESI Legacy Imaging Surveys data. CatBoost is used for point-estimate photometric-redshift regression, and FlexZBoost is used for full redshift-PDF estimation. The workflow achieves robust performance, with a normalised median absolute deviation of 0.036 and an outlier fraction of 5.6% on the test sample. Applying the trained model to the MGQPC catalogue, we identify 185 high-probability quasar pair candidates based on photometric-redshift consistency. Among them, 20 systems have been subsequently confirmed as genuine physical pairs by independent spectroscopic observations. The resulting MGQPC photometric-redshift catalogue provides a useful resource for future spectroscopic follow-up of quasar pairs and dual supermassive black holes.

astro-ph.GA↗