arXiv subjects

Yanhui Wang

Publications and source records attributed to Yanhui Wang.

16 recordsLinked to original sources

arXiv2026

Translation as a Bridging Action: Transferring Manipulation Skills from Humans to Robots

We study whether we can learn novel manipulation skills from human actions to a bi-manual robot with parallel grippers. Human action data is cheap, abundant, and diverse, making it one of the most promising resources for scaling up robot learning. Yet transferring skills from humans to robots remains hard: most prior work treats humans as just another bi-manual 6DoF embodiment, where hand-pose estimates are noisy and the contact patterns of human fingers differ fundamentally from those of a parallel gripper. We argue that learning rotation-inclusive action signals from human data is therefore sub-optimal, and instead propose a bridging action representation: the relative wrist translation within the initial head-camera frame, an action space shared by humans and robots. To handle the potential absence of certain action components in different embodiments, we build a $\pi_0$-like vision-language-action model with interleaved action tokens and attention masking. On a suite of novel bi-manual manipulation tasks, our bridging action transfers human manipulation knowledge to robots far more effectively than noisy 6DoF human actions and scales with the amount of human data.

cs.RO↗

arXiv2026

Left Ehresmann monoids with a proper basis

Left Ehresmann monoids, and their two-sided counterpart of Ehresmann monoids, were so named by Lawson, who elucidated their connection to the work of Ehresmann in differential geometry. This article is dedicated to building a theory for left Ehresmann monoids inspired by that for inverse semigroups; in order to do so we must develop substantially different ideas and techniques. It is known that every left Ehresmann monoid has a cover, that is, a projection separating preimage, of the form $\mathcal{P}_{\ell}(T,X)$, where $\mathcal{P}_{\ell}(T,X)$ is a left Ehresmann monoid constructed from a monoid $T$ and an order-preserving action of $T$ on a semilattice $X$ with identity. We introduce the notion of a proper basis, and show that $\mathcal{P}_{\ell}(T,X)$, and consequently any free left Ehresmann monoid, possesses a proper basis. We show that any left Ehresmann monoid with a proper basis displays properties close to those of two-sided Ehresmann monoids. Next, we exhibit a class of subsemigroups $\mathcal{Q}_{\ell}(T,X,Y)$ (properly, biunary monoid subsemigroups) of the monoids $\mathcal{P}_{\ell}(T,X)$, which are also left Ehresmann with a proper basis. We prove that any left Ehresmann monoid with a proper basis is isomorphic to some $\mathcal{Q}_{\ell}(T,X,Y)$. Our results can be regarded as being analogous to those for proper inverse semigroups, due to McAlister and O'Carroll, the $\mathcal{Q}_{\ell}(T,X,Y)$ playing the role of the $P$-semigroups and the $\mathcal{P}_{\ell}(T,X)$ the role of the semidirect products of a semilattice by a group. In the process of proving our main theorems we present a globalisation result for an order-preserving partial action of a monoid on a partially ordered set or semilattice.

math.RA↗

arXiv2025

On the diameter and girth of zero-divisor graphs of inverse semigroups

Let $S$ be an inverse semigroup with zero and let $Z(S)^\times$ be its set of non-zero divisors with respect to the natural partial order $\le $ on $S$, that is, $a \in Z(S)^\times $ if there exists $b\in S\setminus\{0\}$ with $\omega(a, b) = \{c \in S: c \leq a\ \mbox{and}\ c \leq b\}=\{0\}$. The set $Z(S)^\times$ makes up the vertices of the corresponding {\it zero-divisor graph} $\Gamma (S)$, with two distinct vertices $a, b$ forming an edge if $\omega(a, b)=\{0\}$. We characterize {\it zero-divisor graphs} of inverse semigroups in terms of their diameter and girth. We also classify inverse semigroups without zero by building a connection between the diameter (girth) and the least group congruence $\sigma$ on an inverse semigroup without zero. Finally, we give a description of the diameter and girth of graph inverse semigoups $I(G)$ in terms of the set of vertices and the set of edges of a graph $G$.

math.GR↗

arXiv2024

Network right * abundant semigroups

We introduce the class of network right $*$-abundant semigroups. These are based on networks that extend the notion of a directed graph. This class properly contains the class of graph inverse semigroups. We investigate the structure of network right $*$-abundant semigroups. We show that two network right $*$-abundant semigroups are isomorphic if and only if the underlying networks are isomorphic.

math.GR↗

arXiv2024

Panacea+: Panoramic and Controllable Video Generation for Autonomous Driving

The field of autonomous driving increasingly demands high-quality annotated video training data. In this paper, we propose Panacea+, a powerful and universally applicable framework for generating video data in driving scenes. Built upon the foundation of our previous work, Panacea, Panacea+ adopts a multi-view appearance noise prior mechanism and a super-resolution module for enhanced consistency and increased resolution. Extensive experiments show that the generated video samples from Panacea+ greatly benefit a wide range of tasks on different datasets, including 3D object tracking, 3D object detection, and lane detection tasks on the nuScenes and Argoverse 2 dataset. These results strongly prove Panacea+ to be a valuable data generation framework for autonomous driving.

cs.CV↗

arXiv2024

CCEdit: Creative and Controllable Video Editing via Diffusion Models

In this paper, we present CCEdit, a versatile generative video editing framework based on diffusion models. Our approach employs a novel trident network structure that separates structure and appearance control, ensuring precise and creative editing capabilities. Utilizing the foundational ControlNet architecture, we maintain the structural integrity of the video during editing. The incorporation of an additional appearance branch enables users to exert fine-grained control over the edited key frame. These two side branches seamlessly integrate into the main branch, which is constructed upon existing text-to-image (T2I) generation models, through learnable temporal layers. The versatility of our framework is demonstrated through a diverse range of choices in both structure representations and personalized T2I models, as well as the option to provide the edited key frame. To facilitate comprehensive evaluation, we introduce the BalanceCC benchmark dataset, comprising 100 videos and 4 target prompts for each video. Our extensive user studies compare CCEdit with eight state-of-the-art video editing methods. The outcomes demonstrate CCEdit's substantial superiority over all other methods.

cs.CV↗

arXiv2024

How to Understand Named Entities: Using Common Sense for News Captioning

News captioning aims to describe an image with its news article body as input. It greatly relies on a set of detected named entities, including real-world people, organizations, and places. This paper exploits commonsense knowledge to understand named entities for news captioning. By ``understand'', we mean correlating the news content with common sense in the wild, which helps an agent to 1) distinguish semantically similar named entities and 2) describe named entities using words outside of training corpora. Our approach consists of three modules: (a) Filter Module aims to clarify the common sense concerning a named entity from two aspects: what does it mean? and what is it related to?, which divide the common sense into explanatory knowledge and relevant knowledge, respectively. (b) Distinguish Module aggregates explanatory knowledge from node-degree, dependency, and distinguish three aspects to distinguish semantically similar named entities. (c) Enrich Module attaches relevant knowledge to named entities to enrich the entity description by commonsense information (e.g., identity and social position). Finally, the probability distributions from both modules are integrated to generate the news captions. Extensive experiments on two challenging datasets (i.e., GoodNews and NYTimes) demonstrate the superiority of our method. Ablation studies and visualization further validate its effectiveness in understanding named entities.

cs.CL↗

arXiv2023

MicroCinema: A Divide-and-Conquer Approach for Text-to-Video Generation

We present MicroCinema, a straightforward yet effective framework for high-quality and coherent text-to-video generation. Unlike existing approaches that align text prompts with video directly, MicroCinema introduces a Divide-and-Conquer strategy which divides the text-to-video into a two-stage process: text-to-image generation and image\&text-to-video generation. This strategy offers two significant advantages. a) It allows us to take full advantage of the recent advances in text-to-image models, such as Stable Diffusion, Midjourney, and DALLE, to generate photorealistic and highly detailed images. b) Leveraging the generated image, the model can allocate less focus to fine-grained appearance details, prioritizing the efficient learning of motion dynamics. To implement this strategy effectively, we introduce two core designs. First, we propose the Appearance Injection Network, enhancing the preservation of the appearance of the given image. Second, we introduce the Appearance Noise Prior, a novel mechanism aimed at maintaining the capabilities of pre-trained 2D diffusion models. These design elements empower MicroCinema to generate high-quality videos with precise motion, guided by the provided text prompts. Extensive experiments demonstrate the superiority of the proposed framework. Concretely, MicroCinema achieves SOTA zero-shot FVD of 342.86 on UCF-101 and 377.40 on MSR-VTT. See https://wangyanhui666.github.io/MicroCinema.github.io/ for video samples.

cs.CV↗

arXiv2023

ART$\boldsymbol{\cdot}$V: Auto-Regressive Text-to-Video Generation with Diffusion Models

We present ART$\boldsymbol{\cdot}$V, an efficient framework for auto-regressive video generation with diffusion models. Unlike existing methods that generate entire videos in one-shot, ART$\boldsymbol{\cdot}$V generates a single frame at a time, conditioned on the previous ones. The framework offers three distinct advantages. First, it only learns simple continual motions between adjacent frames, therefore avoiding modeling complex long-range motions that require huge training data. Second, it preserves the high-fidelity generation ability of the pre-trained image diffusion models by making only minimal network modifications. Third, it can generate arbitrarily long videos conditioned on a variety of prompts such as text, image or their combinations, making it highly versatile and flexible. To combat the common drifting issue in AR models, we propose masked diffusion model which implicitly learns which information can be drawn from reference images rather than network predictions, in order to reduce the risk of generating inconsistent appearances that cause drifting. Moreover, we further enhance generation coherence by conditioning it on the initial frame, which typically contains minimal noise. This is particularly useful for long video generation. When trained for only two weeks on four GPUs, ART$\boldsymbol{\cdot}$V already can generate videos with natural motions, rich details and a high level of aesthetic quality. Besides, it enables various appealing applications, e.g., composing a long video from multiple text prompts.

cs.CV↗

arXiv2023

Panacea: Panoramic and Controllable Video Generation for Autonomous Driving

The field of autonomous driving increasingly demands high-quality annotated training data. In this paper, we propose Panacea, an innovative approach to generate panoramic and controllable videos in driving scenarios, capable of yielding an unlimited numbers of diverse, annotated samples pivotal for autonomous driving advancements. Panacea addresses two critical challenges: 'Consistency' and 'Controllability.' Consistency ensures temporal and cross-view coherence, while Controllability ensures the alignment of generated content with corresponding annotations. Our approach integrates a novel 4D attention and a two-stage generation pipeline to maintain coherence, supplemented by the ControlNet framework for meticulous control by the Bird's-Eye-View (BEV) layouts. Extensive qualitative and quantitative evaluations of Panacea on the nuScenes dataset prove its effectiveness in generating high-quality multi-view driving-scene videos. This work notably propels the field of autonomous driving by effectively augmenting the training dataset used for advanced BEV perception techniques.

cs.CV↗

arXiv2023

FinEval: A Chinese Financial Domain Knowledge Evaluation Benchmark for Large Language Models

Large language models have demonstrated outstanding performance in various natural language processing tasks, but their security capabilities in the financial domain have not been explored, and their performance on complex tasks like financial agent remains unknown. This paper presents FinEval, a benchmark designed to evaluate LLMs' financial domain knowledge and practical abilities. The dataset contains 8,351 questions categorized into four different key areas: Financial Academic Knowledge, Financial Industry Knowledge, Financial Security Knowledge, and Financial Agent. Financial Academic Knowledge comprises 4,661 multiple-choice questions spanning 34 subjects such as finance and economics. Financial Industry Knowledge contains 1,434 questions covering practical scenarios like investment research. Financial Security Knowledge assesses models through 1,640 questions on topics like application security and cryptography. Financial Agent evaluates tool usage and complex reasoning with 616 questions. FinEval has multiple evaluation settings, including zero-shot, five-shot with chain-of-thought, and assesses model performance using objective and subjective criteria. Our results show that Claude 3.5-Sonnet achieves the highest weighted average score of 72.9 across all financial domain categories under zero-shot setting. Our work provides a comprehensive benchmark closely aligned with Chinese financial domain.

cs.CL↗

arXiv2022

Lyapunov exponent, universality and phase transition for products of random matrices

Products of $M$ i.i.d. random matrices of size $N \times N$ are related to classical limit theorems in probability theory ($N=1$ and large $M$), to Lyapunov exponents in dynamical systems (finite $N$ and large $M$), and to universality in random matrix theory (finite $M$ and large $N$). Under the two different limits of $M \to \infty$ and $N \to \infty$, the local singular value statistics display Gaussian and random matrix theory universality, respectively. However, it is unclear what happens if both $M$ and $N$ go to infinity. This problem, proposed by Akemann, Burda, Kieburg \cite{Akemann-Burda-Kieburg14} and Deift \cite{Deift17}, lies at the heart of understanding both kinds of universal limits. In the case of complex Gaussian random matrices, we prove that there exists a crossover phenomenon as the relative ratio of $M$ and $N$ changes from $0$ to $\infty$: sine and Airy kernels from the Gaussian Unitary Ensemble (GUE) when $M/N \to 0$, Gaussian fluctuation when $M/N \to \infty$, and new critical phenomena when $M/N \to γ\in (0,\infty)$. Accordingly, we further prove that the largest singular value undergoes a phase transition between the Gaussian and GUE Tracy-Widom distributions.

math.PR↗

arXiv2022

Quasi-semilattices on networks

This paper introduces the tensor representation of a network, here tensors are the primitive structures of the network. In view of tensor chains, two binary operations on tensor sets are defined: chain addition and reducing. Based on the reducing operation, the tensor chain representation of subnetworks of a network is given, and it is proved that all connected subnetworks of a network (here refers to the tensor chain generated by primitive structures) form a quasi-semilattice with respect to reducing, namely {\it network quasi-semilattices}. Here, quasi-semilattices refer to algebraic systems that are idempotent commutative and do not satisfy the association law. Then, we discuss the subalgebra structures of the network quasi-semilattice in terms of two equivalent relations $σ$ and $δ$. $δ$ is a congruence. Each $δ$-class forms a semilattice with respect to reducing, that is, an idempotent commutative semigroup, and also each $δ$-class has an order structure with the maximum element and minimum elements. Here, the minimum elements correspond to the spanning tree in graph theory. Finally, we discuss how three path algebras: graph inverse semigroups, Leavitt path algebra and Cuntz-Krieger graph $C^*$-algebra are constructed in terms of tensors with respect to chain-addition.

math.RA↗

arXiv2019

Phase transitions for infinite products of large non-Hermitian random matrices

Products of $M$ i.i.d. non-Hermitian random matrices of size $N \times N$ relate Gaussian fluctuation of Lyapunov and stability exponents in dynamical systems (finite $N$ and large $M$) to local eigenvalue universality in random matrix theory (finite $M$ and large $N$). The remaining task is to study local eigenvalue statistics as $M$ and $N$ tend to infinity simultaneously, which lies at the heart of understanding two kinds of universal patterns. For products of i.i.d. complex Ginibre matrices, truncated unitary matrices and spherical ensembles, as $M+N\to \infty$ we prove that local statistics undergoes a transition when the relative ratio $M/N$ changes from $0$ to $\infty$: Ginibre statistics when $M/N \to 0$, normality when $M/N\to \infty$, and new critical phenomena when $M/N\to γ\in (0, \infty)$.

math.PR↗

arXiv2019

Singular Value Statistics for the Spiked Elliptic Ginibre Ensemble

The complex elliptic Ginibre ensemble with coupling $τ$ is a complex Gaussian matrix interpolating between the Gaussian Unitary Ensemble (GUE) and the Ginibre ensemble. It has been known for some time that its eigenvalues form a determinantal point process in the complex plane. A recent result of Kanazawa and Kieburg (arXiv:1804.03985) shows that the singular values form a Pfaffian point process. In this paper we turn to consider an extended elliptic Ginibre ensemble, which connects the GUE and the spiked Wishart matrix, and prove that the singular values still build a Pfaffian point process with correlation kernels expressed by contour integral representations. As $τ$ tends to 1 at a certain critical rate, we prove that the limiting distribution of the largest singular value is described as a new Fredholm Pfaffian series, which connects two distributions $F_{\mathrm{GUE}}$ and $F^{2}_{\mathrm{GUE}}$ where $F_{\mathrm{GUE}}$ is the GUE Tracy-Widom distribution. For fixed $τ$, we prove the Baik-Ben Arous-Péché transition of the largest singular value and the sine kernel in the bulk. We also observe a crossover phenomenon at the origin when $τ$ tends to 1 at another critical rate.

math.PR↗

arXiv2015

Universality for products of random matrices I: Ginibre and truncated unitary cases

Recently, the joint probability density functions of complex eigenvalues for products of independent complex Ginibre matrices have been explicitly derived as determinantal point processes. We express truncated series coming from the correlation kernels as multivariate integrals with singularity and investigate saddle point method for such a type of integrals. As an application, we prove that the eigenvalue correlation functions have the same scaling limits as those of the single complex Ginibre ensemble, both in the bulk and at the edge of the spectrum. We also prove that the similar results hold true for products of independent truncated unitary matrices.

math.PR↗