Searcharxiv⌕ Search

arXiv subjects

Zhihan Wang

Publications and source records attributed to Zhihan Wang.

At least 19 recordsLinked to original sources

Auditing Multi-Proxy Explanations of Web User Preferences: Identification Boundaries and Abstention

Web systems infer user preferences from behavior, ratings, text, and LLM-generated profiles, yet explaining a proxy may reveal its measurement process rather than the preference of interest. We develop an audit framework distinguishing point identification, partial identification, and abstention relative to an explicit reference distribution. Under feature-dependent, arbitrarily signed proxy bias, a strict majority of uncontaminated proxy attributions lets the coordinatewise median recover latent-attribution support and direction and, with externally justified positive scales, magnitude. We prove this threshold sharp using response-valid indistinguishable worlds, characterize sharp minimax error and exact compatible-target sets under approximate cleanliness and scale uncertainty, and propagate clean-proxy attribution error. A frozen known-truth benchmark spans two response families, four proxy counts, all contamination counts, three bias strengths, and eleven methods, with a focal comparison confirmed on twenty unseen seeds. Under correct budgets, a direct attribution-tensor audit achieved 100% coverage, 95.40%-97.05% active-driver sign decisions, and no false signs. In a temporally separated Amazon study, median aggregation reduced comparator-relative conditional sign error for all four targets; six of eight prespecified target-metric contrasts had Bonferroni intervals excluding zero. A controlled real-covariate perturbation favored the median on the prespecified focal metric for all four targets. Compatible sets showed that the current proxy configuration rarely licensed strict-sign conclusions. Weak comparator calibration limits the Amazon evidence to fitted-surface alignment rather than latent-preference recovery.

cs.LG↗

When Category Means Move: A Paired-Support Audit of Recommendation Logs with Random Insertions

Content-category dashboards summarize production recommendation logs, although the serving policy selects the users, items, and contexts behind each mean. Using KuaiRand random-insertion impressions, we audit these summaries on the same 12,472 users and 5,486 videos in both logs. Standard means have absolute/centered mean absolute error (MAE) 0.1935/0.0491 and rank correlation 0.7833. User demeaning plus a random grand-mean anchor reduces absolute MAE by 80.0%, but centered MAE by only 21.0%, without reliably improving rank. A complete-support ledger allocates weighted squared error mainly to within-video residual (72.41%) and video mix (23.90%), with 3.65% from random support. In a frozen later test, early category calibration reaches MAE 0.0251 versus 0.0531 for scalar calibration; their direct difference is -0.0280 [-0.0296, -0.0172]. Decision evidence is cutoff-specific: the nominal top-10 regret change for row-weighted demeaning versus naive is -0.0062 [-0.0084, -0.0003], while top-5 and top-20 intervals include zero. The six-contrast family-wise top-10 interval [-0.0136, 0.0013] includes zero. An outcome-free disjoint randomized-reference split also leaves all six nominal contrasts covering zero. A known-truth stress test shows incomplete improvement under a mean-additive positive control and failure under item mix and weak overlap. All decompositions are descriptive and each result is conditional on its overlap cohort. KuaiRand intervals quantify user-cluster variation conditional on the realized catalog and calendar; we claim neither intrinsic category effects nor a universally valid correction.

cs.IR↗

Observing minimal submanifolds with isolated singularities on open sets of metrics

We develop a perturbative theory for minimal submanifolds with only (strongly) isolated singularities, with respect to varying background Riemannian metrics. As an application, we obtain a general "sporadic persistence" criterion, thus naturally complementing the non-persistence theorem by the authors. In particular, such a criterion is satisfied by the spherical suspension in $\mathbb{S}^{3m+2}$ of the Veronese embedding of $\mathbb{K}\mathbb{P}^2$ for $\mathbb{K}=\mathbb{R},\mathbb{C},\mathbb{H},\mathbb{O}$ (the normed division algebras) in the round spheres of dimension $3m+1=4,7,13,25$ respectively, thereby providing families of continuously varying $(2m+1)$-dimensional stationary varifolds with isolated singularities that are observable on an open set of metrics. By our previous results, these phenomena cannot possibly occur in the codimension one scenario.

math.DG↗

Morse index characterization of the Veronese projective plane

Following Simons' seminal work in 1968, every closed, connected, embedded minimal surface in the four-sphere that is not totally geodesic has Morse index at least five. In this paper, we prove that equality holds if and only if the surface is congruent to the Veronese projective plane. Relying on recent advances by Bernstein and Ketover, this characterization settles Kusner's 1996 conjecture on the minimizers of the Willmore functional in dimension four for surfaces whose normal bundle has nonzero Euler number. This is always the case, in particular, for embeddings of nonorientable surfaces whose orientation double cover has even genus.

math.DG↗

JaxAHT: A JAX-Based Library for Ad Hoc Teamwork

As autonomous agents are increasingly deployed, they must be able to coordinate with previously unseen agents, such as humans. This challenge has been formalized as Ad Hoc Teamwork (AHT), and has attracted a substantial body of research. However, progress has been hindered by the prohibitive computational cost of the AHT research lifecycle, lack of standardized benchmark implementations, and the absence of a diverse, validated evaluation teammate suite. This work introduces JaxAHT, the first open-source, JAX-based library designed to accelerate and standardize the AHT research lifecycle. Leveraging JAX's hardware acceleration and massive parallelization capabilities, JaxAHT provides a unified framework for teammate generation, ego agent training, and evaluation against unseen teammates, achieving approximately 95x wall-clock speedup over existing PyTorch counterparts. Along with the library, we contribute a diverse suite of evaluation teammates across Level-Based Foraging, Overcooked, and Hanabi. Using JaxAHT, we conduct a large-scale, compute-controlled benchmark study of AHT algorithms, finding that no algorithm consistently performs best, that agent modeling primarily offers benefits in role-based scenarios, and that best response returns are underestimated by reinforcement learning from scratch.

cs.AI↗

Min-max theory and minimal surfaces with prescribed genus

We establish a general min-max type theorem that produces minimal surfaces with prescribed genus in 3-manifolds with positive Ricci curvature. An important intermediate step is to show that, in a generic metric with positive Ricci curvature, any family of surfaces with possibly finitely many singularities can be deformed into a certain topologically optimal family. Results in this paper will be crucial to our program on the construction of multiple minimal surfaces with prescribed genus in 3-spheres via topological methods.

math.DG↗

FunnelCausalNet: Funnel-aware Joint Conversion-Revenue Uplift for Multi-tier Coupon Allocation

Coupon campaigns seek to lift both conversion and revenue, but gross merchandise value (GMV) follows a deterministic funnel from conversion to conditional order value and is zero-inflated and heavy-tailed. We propose FunnelCausalNet, an uplift estimator coupling a binary conversion head with a nonnegative conditional-value head through $μ_{\mathrm{gmv}}=μ_{\mathrm{conv}}μ_{\mathrm{val}}$. Under explicit RCT, support, rate-gap, and cross-head covariance-control assumptions, an idealized leading-order MSE comparison identifies a regime in which funnel composition can reduce pointwise variance; this is a heuristic, not a guarantee for the shared-representation neural model. The estimator is paired with marginal split-conformal CATE summaries, combined through a Bonferroni union as audit bands, and a Lagrangian budgeted allocator using RCT-anchored estimates for subsidy-aware ROI accounting. On semi-synthetic multi-tier Criteo-MT7, FunnelCausalNet's mean AUUC_GMV is within one seed standard deviation of the leading feature-interaction baseline among eleven baselines, while a controlled ablation reduces GMV effect error versus direct GMV regression by 18--48% across tested zero-inflation regimes. On de-identified industrial Hotel-Coupon RCT logs with about 4.9 million hold-out exposure records per seed, expected-outcome evaluation sweeps full LP frontiers; FunnelCausalNet has the best seed-averaged mean DeltaROI at all seven correlated anchors from 10% to 60%, which we treat as descriptive frontier consistency rather than independent significance. On sparse binary-spend public benchmarks, revenue-focused rankers can dominate uplift-curve proxies, defining an explicit regime boundary.

cs.LG↗

Learning Agile Striker Skills for Humanoid Soccer Robots from Noisy Sensory Input

Learning fast and robust ball-kicking skills is a critical capability for humanoid soccer robots, yet it remains a challenging problem due to the need for rapid leg swings, postural stability on a single support foot, and robustness under noisy sensory input and external perturbations (e.g., opponents). This paper presents a reinforcement learning (RL)-based system that enables humanoid robots to execute robust continual ball-kicking with adaptability to different ball-goal configurations. The system extends a typical teacher-student training framework -- in which a "teacher" policy is trained with ground truth state information and the "student" learns to mimic it with noisy, imperfect sensing -- by including four training stages: (1) long-distance ball chasing (teacher); (2) directional kicking (teacher); (3) teacher policy distillation (student); and (4) student adaptation and refinement (student). Key design elements -- including tailored reward functions, realistic noise modeling, and online constrained RL for adaptation and refinement -- are critical for closing the sim-to-real gap and sustaining performance under perceptual uncertainty. Extensive evaluations in both simulation and on a real robot demonstrate strong kicking accuracy and goal-scoring success across diverse ball-goal configurations. Ablation studies further highlight the necessity of the constrained RL, noise modeling, and the adaptation stage. This work presents a system for learning robust continual humanoid ball-kicking under imperfect perception, establishing a benchmark task for visuomotor skill learning in humanoid whole-body control.

cs.RO↗

An enumerative min-max theorem for minimal surfaces

We prove an enumerative min-max theorem that relates the number of genus g minimal surfaces in 3-manifolds of positive Ricci curvature to topological properties of the set of embedded surfaces of genus $\leq g$, possibly with finitely many singularities. This completes a central component of our program of using topological methods to enumerating minimal surfaces with prescribed genus. As an application, we show that every 3-sphere of positive Ricci curvature contains at least 4 embedded minimal surfaces of genus 2.

math.DG↗

Evaluating Generalization Capabilities of LLM-Based Agents in Mixed-Motive Scenarios Using Concordia

Large Language Model (LLM) agents have demonstrated impressive capabilities for social interaction and are increasingly being deployed in situations where they might engage with both human and artificial agents. These interactions represent a critical frontier for LLM-based agents, yet existing evaluation methods fail to measure how well these capabilities generalize to novel social situations. In this paper, we introduce a method for evaluating the ability of LLM-based agents to cooperate in zero-shot, mixed-motive environments using Concordia, a natural language multi-agent simulation environment. Our method measures general cooperative intelligence by testing an agent's ability to identify and exploit opportunities for mutual gain across diverse partners and contexts. We present empirical results from the NeurIPS 2024 Concordia Contest, where agents were evaluated on their ability to achieve mutual gains across a suite of diverse scenarios ranging from negotiation to collective action problems. Our findings reveal significant gaps between current agent capabilities and the robust generalization required for reliable cooperation, particularly in scenarios demanding persuasion and norm enforcement.

cs.AI↗

Regularity of cylindrical singular sets of mean curvature flow

In this paper, we study the $k$-cylindrical singular set of mean curvature flow in $\mathbb R^{n+1}$ for each $1\leq k\leq n-1$. We prove that they are locally contained in a $k$-dimensional $C^{2,α}$-submanifold after removing some lower-dimensional parts. Moreover, if the $k$-cylindrical singular set is a $k$-submanifold, then its curvature is determined by the asymptotic profile of the flow at these singularities. As a byproduct, we provide a detailed asymptotic profile and graphical radius estimate at these singularities. The proof is based on a new $L^2$-distance non-concentration property that we introduced in [SWX25], modified into a relative version that allows us to modulo those low eigenmodes that are not decaying fast enough and do not contribute to the curvature of the singular set.

math.DG↗

On mean curvature flow translators with prescribed ends

Given a smooth closed embedded self-shrinker $S$ with index $I$ in $\mathbb{R}^{n}$, we construct an $I$-dimensional family of complete translators polynomially asymptotic to $S\times\mathbb{R}$ at infinity, which answers a long-standing question by Ilmanen. We further prove that $\mathbb{R}^{n+1}$ can be decomposed in many ways into a one-parameter family of closed sets $\coprod_{a\in \mathbb{R}} T_a$, and each closed set $T_a$ contains a complete translator asymptotic to $S\times\mathbb{R}$ at infinity. If the closed set $T_a$ fattens, namely it has nonempty interior, then there are at least two translators asymptotic to each other at an exponential rate, which can be viewed as a kind of nonuniqueness. We show that this fattening phenomenon is non-generic but indeed happens.

math.DG↗

Generic regularity for minimizing hypersurfaces in dimension 11

We prove that area-minimizing hypersurfaces are generically smooth in ambient dimension $11$ in the context of the Plateau problem and of area minimization in integral homology. For higher ambient dimensions, $n+1 \geq 12$, we prove in the same two contexts that area-minimizing hypersurfaces have at most an $n-10-ε_n$ dimensional singular set after an arbitrarily $C^\infty$-small perturbation of the Plateau boundary or the ambient Riemannian metric, respectively.

math.DG↗

Reinforcement Learning Within the Classical Robotics Stack: A Case Study in Robot Soccer

Robot decision-making in partially observable, real-time, dynamic, and multi-agent environments remains a difficult and unsolved challenge. Model-free reinforcement learning (RL) is a promising approach to learning decision-making in such domains, however, end-to-end RL in complex environments is often intractable. To address this challenge in the RoboCup Standard Platform League (SPL) domain, we developed a novel architecture integrating RL within a classical robotics stack, while employing a multi-fidelity sim2real approach and decomposing behavior into learned sub-behaviors with heuristic selection. Our architecture led to victory in the 2024 RoboCup SPL Challenge Shield Division. In this work, we fully describe our system's architecture and empirically analyze key design decisions that contributed to its success. Our approach demonstrates how RL-based behaviors can be integrated into complete robot behavior architectures.

cs.RO↗

Non-persistence of strongly isolated singularities, and geometric applications

We obtain a generic regularity result for stationary integral $n$-varifolds with only strongly isolated singularities inside $N$-dimensional Riemannian manifolds, in absence of any restriction on the dimension ($n\geq 2$) and codimension. As a special case, we prove that for any $n\geq 2$ and any compact $(n+1)$-dimensional manifold $M$ the following holds: for a generic choice of the background metric $g$ all stationary integral $n$-varifolds in $(M,g)$ will either be entirely smooth or have at least one singular point that is not strongly isolated. In other words, only ``more complicated'' singularities may possibly persist. This implies, for instance, a generic finiteness result for the class of all closed minimal hypersurfaces of area at most $4π^2-\varepsilon$ (for any $\varepsilon>0$) in nearly round four-spheres: we can thus give precise answers, in the negative, to the questions of persistence of the Clifford football and of Hsiang's hyperspheres in nearly round metrics. The aforementioned main regularity result is achieved as a consequence of the fine analysis of the Fredholm index of the Jacobi operator for such varifolds: we prove on the one hand an exact formula relating that number to the Morse indices of the conical links at the singular points, while on the other hand we show that the same number is non-negative for all such varifolds if the ambient metric is generic.

math.DG↗

Passing through nondegenerate singularities in mean curvature flows

In this paper, we study the properties of nondegenerate cylindrical singularities of mean curvature flow. We prove they are isolated in spacetime and provide a complete description of the geometry and topology change of the flow passing through the singularities. Particularly, the topology change agrees with the level sets change near a critical point of a Morse function, which is the same as performing surgery. The proof is based on a new $L^2$-distance monotonicity formula, which allows us to derive a discrete almost monotonicity of the ``decay order", a discrete mean curvature flow analog to Almgren's frequency function.

math.DG↗

Synergistic PET/CT Reconstruction Using a Joint Generative Model

We propose in this work a framework for synergistic positron emission tomography (PET)/computed tomography (CT) reconstruction using a joint generative model as a penalty. We use a synergistic penalty function that promotes PET/CT pairs that are likely to occur together. The synergistic penalty function is based on a generative model, namely $β$-variational autoencoder ($β$-VAE). The model generates a PET/CT image pair from the same latent variable which contains the information that is shared between the two modalities. This sharing of inter-modal information can help reduce noise during reconstruction. Our result shows that our method was able to utilize the information between two modalities. The proposed method was able to outperform individually reconstructed images of PET (i.e., by maximum likelihood expectation maximization (MLEM)) and CT (i.e., by weighted least squares (WLS)) in terms of peak signal-to-noise ratio (PSNR). Future work will focus on optimizing the parameters of the $β$-VAE network and further exploration of other generative network models.

physics.med-ph↗