arXiv subjects

Xin Liu

Publications and source records attributed to Xin Liu.

At least 73 records · Page 4Linked to original sources

arXiv2026

SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents

Generative Engine Optimization (GEO) lets content owners rewrite web content to increase their visibility in generative systems. In recommendation agents, this creates a risk that sources controlled by sellers make flawed products appear better supported than they are. We study this risk at the generation stage by asking whether recommendation agents continue to make decisions that align with user utility when these sources are rewritten for GEO. To make this question measurable, we construct SafeGEO, an evaluation suite with 22 GEO attack variants across 600 recommendation cases. We empirically show that GEO attacks can promote flawed target products: they increase the rate at which such flawed products enter the recommendation set by up to 83.2 percentage points (pp). We further study whether agent-side design choices can mitigate this risk and show that simple defenses reduce harmful target promotion by up to 39.2 pp. These gains are substantial but do not restore the no-GEO performance, showing that GEO remains a serious risk despite mitigation.

cs.IR↗

arXiv2026

TransfHAR: Self-Supervised Wrist Representations for On-Demand Activity Recognition

Fine-grained wrist activity recognition can support applications such as procedural step guidance and context-aware assistance, yet acquiring labeled data for every new task, user, and activity granularity remains a bottleneck. We present TransfHAR, a self-supervised wrist IMU framework for on-demand, fine-grained activity recognition by learning transferable motion priors from global, unlabeled activities. We show that self-supervised pretraining on coarse wrist IMU activities (e.g., sitting, walking, exercise) learns motion structure rich enough to transfer to fine-grained manipulative, gestural, and procedural activities (e.g., snapping, stirring, waving) that are absent from pretraining. We implement TransfHAR as a real-time smartwatch application that lets users define and expand their own activity set for personalized recognition from only a few demonstrations. Across three offline cross-dataset evaluations, TransfHAR matches or exceeds fully supervised baselines that use complete label sets with equal or additional sensor channels, by 6.2 balanced-accuracy points on average. In an in-lab study with 10 participants each performing seven novel wrist activities, TransfHAR reaches 86.7% balanced accuracy across participants with five examples per class and 90.4% when updated from a single one-minute recording per class. These results indicate that broad self-supervised wrist pretraining provides an effective foundation for on-demand fine-grained activity recognition.

cs.LG↗

arXiv2026

APEx: Distillation of Agent Procedural Experience for Adaptive Deep Research Question Answering

Deep research agents augment large language models with external tools to answer complex, long-horizon questions through multi-turn reasoning. Learning from prior experience is crucial for continual improvement, yet existing methods either retrieve verbose task-specific traces that burden decision-making, or distill procedural skills that remain decoupled from downstream policy adaptation. We propose APEx, a hierarchical experience utilization framework that organizes interaction history into instance-level trajectory memories and category-level procedural skills, and couples them through a closed-loop architecture of Executor, Distiller, and Planner. The three modules are optimized via a three-stage alternating GRPO training paradigm, enabling reward-guided skill distillation rather than fixed-prompt generation. At test time, distilled skills serve as procedural priors for online Planner adaptation through skill-guided test-time reinforcement learning, allowing ground-truth-free self-improvement with skill-alignment regularization to prevent policy drift. Experiments on 7 benchmarks demonstrate that APEx achieves state-of-the-art performance, surpassing GPT-5.4 by 14.7 points and the strongest memory-augmented baseline by 3.0 points.

cs.AI↗

arXiv2026

Beyond Scaling: Calculable Error Bounds of the Power-of-Two-Choices Mean-Field Model in Heavy-Traffic

This paper provides a recipe for deriving calculable approximation errors of mean-field models in heavy-traffic with the focus on the well-known load balancing algorithm---power-of-two-choices (Po2). The recipe combines Stein's method for linearized mean-field models and State Space Concentration (SSC) based on geometric tail bounds. In particular, we divide the state space into two regions, a neighborhood near the mean-field equilibrium and the complement of that. We first use a tail bound to show that the steady-state probability being outside the neighborhood is small. Then, we use a linearized mean-field model and Stein's method to characterize the generator difference, which provides the dominant term of the approximation error. From the dominant term, we are able to obtain an asymptotically-tight bound, a calculable bound, not order-wise scaling results like most results in the literature. Finally, we compare the theoretical bound with numerical evaluations to show the effectiveness of our results. We note that the simulation results show that the bound is valid even for small size systems such as a system with only hundred servers.

cs.PF↗

arXiv2026

Does This Moment Justify the Recommendation? Counterfactual Behavior-Grounded Evidence Retrieval for Personalized Video Recommendation

Personalized video recommendation predicts user preference at the video level, while temporal video grounding localizes query-relevant moments. However, strong localization does not establish whether the retrieved moment constitutes valid evidence for recommending the video to a particular user. We study counterfactual behavior-grounded evidence retrieval, which separates where personalized evidence occurs from whether such evidence exists and evaluates whether model predictions respond consistently when that evidence is replaced. We introduce CBGER-10K, containing 5,000 controlled factual--counterfactual pairs for 3,026 users, where each pair replaces only the focal behavior-supported segment while preserving the user, temporal position, and hard distractors. We further propose CBGER, a compact framework that decouples segment-level localization from video-level evidence estimation and learns both through structured counterfactual supervision. CBGER achieves $0.4432$ MRR, $0.6977$ Pair Accuracy, and $0.6987$ Intervention Consistency across five adapted personalized-highlight and temporal-grounding baselines. Notably, compared with QD-DETR, its MRR improvement is not statistically significant, while Pair Accuracy improves by $11.03$ points. These results show that accurate temporal localization does not necessarily imply reliable personalized evidence existence, motivating explicit evaluation of Whether alongside Where.

cs.CV↗

arXiv2026

A Hybrid Subgradient Method for Nonsmooth Nonconvex Bilevel Optimization

In this paper, we focus on the nonconvex-nonconvex bilevel optimization problem (BLO), where both upper-level and lower-level objectives are nonconvex, with the upper-level problem potentially being nonsmooth. We develop a two-timescale momentum-accelerated subgradient method (TMG) that employs two-timescale stepsizes, and establish its local convergence when initialized within a sufficiently small neighborhood of the feasible region. To develop a globally convergent algorithm for (BLO), we introduce a feasibility restoration scheme (FRG) that drives iterates toward the feasible region. Both (TMG) and (FRG) only require the first-order derivatives of the upper-level and lower-level objective functions, ensuring efficient computations in practice. We then develop a novel hybrid method that alternates between (TMG) and (FRG) and adaptively estimates its hyperparameters. Under mild conditions, we establish the global convergence properties of our proposed algorithm. Preliminary numerical experiments demonstrate the high efficiency and promising potential of our proposed algorithm.

math.OC↗

arXiv2026

Exact Quadratic Penalty Function for Symplectic Eigenvalue Problem

The symplectic eigenvalue problem for symmetric positive-definite (spd) matrices plays a crucial role in various scientific fields, including quantum mechanics and control theory. This paper introduces a trace-penalty minimization method, which transforms the symplectic eigenvalue problem into the unconstrained minimization of the trace-penalty function. We prove the equivalence between the penalty problem and the original constrained optimization problem under mild conditions, in the sense that the second-order stationary points of the trace-penalty function correspond to the solutions of the symplectic eigenvalue problem. Moreover, we develop an algorithm to minimize the trace-penalty function efficiently, which follows the scheme of gradient methods, together with the Barzilai-Borwein (BB) adaptive step-size rule and non-monotone line-search technique. Numerical experiments demonstrate that the proposed algorithm outperforms a wide range of existing methods, such as Riemannian gradient-based methods, in terms of computational efficiency and convergence rate for dense, sparse, and sparse-add-low-rank matrices. These numerical results further demonstrate the great potential of our proposed algorithm, especially in solving large-scale symplectic eigenvalue problems.

math.OC↗

arXiv2026

QUBRIC: Co-Designing Queries and Rubrics for RL Beyond Verifiable Rewards

Rubric-based RL is a promising route for extending reinforcement learning beyond verifiable rewards, yet existing methods optimize rubrics while treating the query distribution as fixed. We identify a structural bottleneck: rubric quality is constrained by query structure. Open-ended queries yield vague rubrics; naively narrowing them introduces fabricated references that no model can verify, so all responses fail and training receives no reward signal. We present QUBRIC, a framework that co-designs queries and rubrics. Teacher-derived key points ground the rewriting of open-ended queries into scenario-based, evaluable questions. Contrastive rubric generation then turns teacher-policy gaps into query-level criteria, and learnability filtering retains only informative query-rubric pairs for GRPO training. QUBRIC achieves a +5.5 point gain on ArenaHard over the SFT baseline. Trained only on instruction-following data, it further transfers to three held-out benchmarks spanning legal, moral, and narrative reasoning (+6.3 points on average), with improvements concentrated in reasoning-related dimensions. These results provide evidence that co-designing queries and rubrics can make rubric-based RL a practical complement to RLVR beyond strictly verifiable tasks.

cs.CL↗

arXiv2026

Global solutions to the Boussinesq equations and the planetary geostrophic approximation

We investigate the three dimensional Boussinesq equations in the planetary geostrophic scale, where the buoyancy/gravity driven flow is near geostrophic balance in the large horizontal spatial scale. By establishing the uniform (with respect to the characteristic number) estimates, we are able to obtain the global weak solutions and the unique global strong solutions to the Boussinesq equations without restriction on the initial data. Furthermore, sending the characteristic number to zero with the established compactness rigorously justifies the planetary geostrophic approximation of the Boussinesq equations. We employ the novel buoyancy flow-free flow decomposition to overcome the singularity caused by the buoyancy/gravity.

math.AP↗

arXiv2026

Simultaneously accounting for the winner's curse and sample structure in Mendelian randomization: bivariate rerandomized inverse variance weighted estimator

The recently developed rerandomized inverse variance weighted (RIVW) estimator provides a simple and efficient framework to break the winner's curse in two-sample Mendelian randomization (MR). However, this method does not account for sample structure (e.g., residual population stratification and sample overlap), a common source of confounding in MR studies. Sample structure can not only distort SNP-exposure and SNP-outcome association estimates but also induce correlation between them, leading exposure-side instrument selection to propagate bias to the outcome side. To address this challenge, we propose the bivariate RIVW (BRIVW) estimator to simultaneously account for the winner's curse and sample structure. The BRIVW estimator extends the RIVW framework by modeling the joint distribution of SNP-exposure and SNP-outcome association estimates, first adjusting their covariance matrix via linkage disequilibrium score regression to account for sample structure and then applying randomized instrument selection and bivariate Rao-Blackwellization to obtain unbiased post-selection association estimates together with an estimator of their covariance matrix. Under mild conditions, we show that the BRIVW estimator is consistent and asymptotically normal. The finite-sample performance of the proposed estimator is evaluated through extensive simulations and real data analyses.

stat.ME↗

arXiv2026

DeepPlanner: Scaling Planning Capability for Deep Research Agents via Advantage Shaping

Large language models (LLMs) augmented with multi-step reasoning and action generation abilities have shown promise in leveraging external tools to tackle complex tasks that require long-horizon planning. However, existing approaches either rely on implicit planning in the reasoning stage or introduce explicit planners without systematically addressing how to optimize the planning stage. As evidence, we observe that under vanilla reinforcement learning (RL), planning tokens exhibit significantly higher entropy than other action tokens, revealing uncertain decision points that remain under-optimized. To address this, we propose DeepPlanner, an end-to-end RL framework that effectively enhances the planning capabilities of deep research agents. Our approach shapes token-level advantage with an entropy-based term to allocate larger updates to high entropy tokens, and selectively upweights sample-level advantages for planning-intensive rollouts. Extensive experiments across seven deep research benchmarks demonstrate that DeepPlanner improves planning quality and achieves state-of-the-art results under a substantially lower training budget.

cs.AI↗

arXiv2026

Zero-Shot Video Restoration and Enhancement with Text-to-Image Latent Diffusion Models and Multi-Modal References

Zero-shot image restoration methods with text-to-image latent diffusion models have achieved great success in universal image restoration tasks without training. However, applying them to video restoration will result in severe temporal flickering. In this paper, we propose a novel framework for zero-shot video restoration and enhancement which uses a text-to-image latent diffusion model and multi-modal references. Through the proposed dual prompt tuning inversion and sampling, the inference time can be reduced to nearly 1/3 of the original. The performance and temporal consistency can be also significantly stregthened. By using the proposed texture-aware video token merging, the temporal correlation between frames can be further utilized to improve the temporal consistency. We futher propose the referenced self-attention and referenced token merging to support image reference. Experimental results demonstrate the superiority of the proposed method in restoring and enhancing temporally consistent videos.

cs.CV↗

arXiv2026

MER 2026: From Discriminative Emotion Recognition to Generative Emotion Understanding

MER2026 marks the fourth edition of the MER series of challenges. The MER series provides valuable data resources to the research community and offers tasks centered on recent research trends, establishing itself as one of the largest challenges in the field. Throughout its history, the focus of MER has shifted from discriminative emotion recognition to generative emotion understanding. Specifically, MER2023 concentrated on discriminative emotion recognition, restricting the emotion recognition scope to fixed basic labels. In MER2024 and MER2025, we transitioned to generative emotion understanding and introduced two new tasks: fine-grained emotion recognition and descriptive emotion analysis, aiming to leverage the extensive vocabulary and multimodal understanding capabilities of Multimodal Large Language Models (MLLMs) to facilitate fine-grained and explainable emotion recognition. Building on this trajectory, MER2026 continues to follow these research trends and contains four tracks: MER-Cross shifts the focus from individual to dyadic interaction scenarios; MER-FG centers on fine-grained emotion recognition; MER-Prefer aims to predict human preferences regarding different emotion descriptions; MER-PS focuses on emotion recognition based on physiological signals. More details regarding the dataset and baselines are available at https://zeroqiaoba.github.io/MER-Challenge/.

cs.HC↗

arXiv2026

Improving O-RADS Risk Stratification from Ultrasound Reports: A Comparative Evaluation of Hybrid versus End-to-End LLM Reasoning Strategies

Background: Automating clinical guideline-based decision-making with large language models (LLMs) remains challenging because of reliability, hallucination, and limited interpretability. We compared the performance of LLMs and reasoning strategies for automated Ovarian-Adnexal Reporting and Data System (O-RADS) classification from free-text pelvic ultrasound reports. Methods: In this retrospective study, consecutive patients with ovarian masses who underwent pelvic ultrasound were included. Eight LLMs were tested with three reasoning strategies: implicit-knowledge end-to-end, rule-informed end-to-end, and a feature-based hybrid architecture that decoupled feature extraction from rule-based classification. The reference standard was O-RADS categorization established by expert consensus. Results: A total of 310 women with 390 ovarian masses were evaluated. The feature-based hybrid architecture using Gemini 3.6 Flash demonstrated the best performance, achieving an accuracy of 99.2% (387 of 390) and almost perfect agreement with the reference standard (weighted kappa = 1.00; 95% CI: 0.99-1.00). Its performance surpassed that of original clinical reports (accuracy, 87.7% [342 of 390]; weighted kappa = 0.94; 95% CI: 0.91-0.96) and end-to-end LLM strategies (accuracy range, 65.6% [256 of 390] to 95.9% [374 of 390]). For structured feature extraction, Gemini 3.6 Flash demonstrated higher overall accuracy than Claude Fable 5 (98.9% vs 97.8%; P < 0.001). The hybrid architecture reduced misclassification errors and mitigated the overstaging tendency observed in original reports. Conclusion: The feature-based hybrid LLM architecture that separates clinical feature extraction from deterministic guideline execution enables highly accurate, reliable, and interpretable automated O-RADS classification, providing a promising approach for standardized, guideline-based clinical decision-making.

cs.AI↗

arXiv2026

Local Well-Posedness of the Inviscid Shallow Water Equations with Surface Tension

We show that the inviscid shallow water equations with gravity and surface tension are locally well-posed in dimensions one and two. In contrast to previous works by Benzoni-Gavage, Danchin, and Descombes, we work only in the real-valued phase space and make use of the symmetry of the system. Our alternative proof for local well-posedness sheds some light on the symmetric structure of the shallow water system and can be used for further study, including designing numerical scheme, dispersion estimate, etc.

math.AP↗

arXiv2026

Degree-Mass Message Passing for Betweenness Ranking in Directed and Undirected Networks

Computing the importance of nodes in networks is a long-standing fundamental problem that has driven extensive study of various centrality measures. A particularly well-known centrality measure is betweenness centrality, whose exact computation becomes prohibitive on large-scale networks. Graph Neural Network (GNN) models have thus been proposed to predict the ranking of nodes by betweenness centrality. However, existing GNN-based methods either have graph-size-dependent parameter counts or are limited to undirected graphs. We propose a lightweight GNN architecture that exploits the empirically observed relationship between betweenness centrality and multi-hop degree mass. This motivates the use of degree masses as size-invariant node features. To improve generalization, we train on synthetic graphs whose degree distributions more closely match those of real-world networks, including directed and undirected scale-free graphs and uniformly directed hyperbolic random graphs. We evaluate our model on 14 real-world networks spanning eight domains, including social, email, and citation networks, across both directed and undirected regimes. The experiments show that our model improves the Kendall $τ_b$ correlation by up to 24.6\% on undirected and 10.9\% on directed graphs, while using 56$\times$ fewer parameters than the lightest competing GNN baseline and achieving competitive inference time, with up to a 24.5$\times$ speedup on selected directed graphs.

cs.LG↗

arXiv2026

Volumetric Radiology AI in the Era of Multimodal Large Language Models

Advances in multimodal large language models (MLLMs) are extending radiological artificial intelligence (AI) beyond task-specific image analysis toward multimodal understanding and reasoning. Volumetric radiology, however, presents a fundamental representational mismatch: clinical interpretation often requires full-volume spatial context and acquisition-dependent quantitative information, whereas current MLLMs are commonly conditioned on selected two-dimensional (2D) images, compressed visual representations, or report-derived text. Reliable volumetric radiology AI therefore requires representations that preserve task-relevant three-dimensional (3D) information and systems that can access, verify, and integrate this information across clinical workflows. In this Review, we examine more than 200 publications through July 2026. We organize the literature around volumetric representation and multimodal understanding at the model level, agentic orchestration at the system level, and their links to clinical applications and evaluation. We review volumetric foundation models, language alignment and compression strategies, and agentic systems that extend MLLMs through planning, tools, memory, and workflow interaction. We distinguish settings in which selected 2D views or report-mediated reasoning may suffice from those that warrant native volumetric modeling. We also introduce a Claim-Design-Validation framework to assess whether technical, workflow, and clinical claims are matched by appropriate design and validation. Across the literature, native volumetric modeling and agentic capabilities depend on the spatial, quantitative, contextual, and workflow requirements of the intended task. Clinical credibility requires faithful volumetric representation, traceable system behavior, claim-aligned validation, and clearly defined human oversight in realistic workflows.

cs.AI↗

arXiv2026

A Tight SDP Relaxation for the Cubic-Quartic Regularization Problem

This paper studies how to compute global minimizers of the cubic-quartic regularization (CQR) problem \[ \min_{s \in \mathbb{R}^n} \quad f_0+g^Ts+\frac{1}{2}s^THs+\fracβ{6}\| s \|^3+ \fracσ{4} \| s\|^4, \] where $f_0$ is a constant, $g$ is an $n$-dimensional vector, $H$ is an $n$-by-$n$ symmetric matrix, and $\| s \|$ denotes the Euclidean norm of $s$. The parameter $σ$ is nonnegative while $β$ can have any sign. The CQR problem arises as a critical subproblem for getting efficient regularization methods for solving unconstrained nonlinear optimization. Its properties are recently well studied by Cartis and Zhu {\it [cubic-quartic regularization models for solving polynomial subproblems in third-order tensor methods, Math. Program, 2025]}. We propose a structured semidefinite programming (SDP) relaxation method for solving the CQR problem globally. The SDP relaxation has only three symmetric positive semidefinite matrix variables of sizes $(n+1)$-by-$(n+1)$, $3$-by-$3$ and $2$-by-$2$ respectively. We show that our SDP relaxation is tight if and only if $\| s^* \| ( β+ 3 σ\| s^* \|) \ge 0$ holds for a global minimizer $s^*$. When $s^* \ne 0$, this aligns with the sufficient global optimality condition $β+ 3 σ\| s^* \| \ge 0$ given by Cartis and Zhu. In particular, if either $β\ge 0$ or $H$ has a nonpositive eigenvalue, then the SDP relaxation is shown to be tight. Second, we show that all nonzero global minimizers have the same Euclidean norm for the tight case. Third, we give an algorithm to detect tightness and to obtain the set of all global minimizers. Numerical experiments demonstrate that our SDP relaxation method is both effective and computationally efficient. This paper gives a polynomial time algorithm for solving the CQR problem globally, under the sufficient global optimality condition.

math.OC↗