arXiv subjects

Yipu Wang

Publications and source records attributed to Yipu Wang.

11 recordsLinked to original sources

arXiv2026

PRM-as-a-Judge 1.5: A Toolkit for Robot Process Assessment

Fine-grained robotic evaluation matters for understanding embodied models, going beyond binary success rates and rule-based process scores. We present PRM-as-a-Judge 1.5, a toolkit for robot process assessment that turns rollout videos into dense progress curves and derives multiple fine metrics. PRM-as-a-Judge 1.5 introduces three metrics, building on version 1.0, that characterize failure-side progress, post-drawdown recovery, and success-side execution quality, helping users understand embodied model capability. Based on the rollout videos from benchmarks, we perform a comprehensive assessment of the embodied models, providing some fine-grained metric results and key findings. We further introduce RoboPulse++ to evaluate the reliability of process reward models (PRM), providing evaluators with a more accurate testing platform. Moreover, we release a user-friendly assessment suite, including the benchmark, metric implementation, and visualization tools, to support reproducible manipulation process evaluation. We call on the community to rethink how robots are evaluated and establish transparent, procedural, and reproducible assessment as a foundation for the next generation of embodied intelligence.

cs.RO↗

arXiv2025

Scaling Up AI-Generated Image Detection with Generator-Aware Prototypes

The pursuit of a universal AI-generated image (AIGI) detector often relies on aggregating data from numerous generators to improve generalization. However, this paper identifies a paradoxical phenomenon we term the Benefit then Conflict dilemma, where detector performance stagnates and eventually degrades as source diversity expands. Our systematic analysis, diagnoses this failure by identifying two core issues: severe data-level heterogeneity, which causes the feature distributions of real and synthetic images to increasingly overlap, and a critical model-level bottleneck from fixed, pretrained encoders that cannot adapt to the rising complexity. To address these challenges, we propose Generator-Aware Prototype Learning (GAPL), a framework that constrain representation with a structured learning paradigm. GAPL learns a compact set of canonical forgery prototypes to create a unified, low-variance feature space, effectively countering data heterogeneity.To resolve the model bottleneck, it employs a two-stage training scheme with Low-Rank Adaptation, enhancing its discriminative power while preserving valuable pretrained knowledge. This approach establishes a more robust and generalizable decision boundary. Through extensive experiments, we demonstrate that GAPL achieves state-of-the-art performance, showing superior detection accuracy across a wide variety of GAN and diffusion-based generators. Code is available at https://github.com/UltraCapture/GAPL

cs.CV↗

arXiv2025

Towards Cross-View Point Correspondence in Vision-Language Models

Cross-view correspondence is a fundamental capability for spatial understanding and embodied AI. However, it is still far from being realized in Vision-Language Models (VLMs), especially in achieving precise point-level correspondence, which is crucial for precise affordance interaction. So we propose the Cross-View Point Correspondence (CVPC) task and CrossPoint-Bench, a comprehensive benchmark with hierarchical design, inspired by the human cognitive process of "perceive", "reason", and "correspond". Our evaluation shows the state-of-the-art models (e.g., Gemini-2.5-Pro) still fall far behind humans, with a gap of over 54.65% in overall accuracy, exposing a challenge in transitioning from coarse-grained judgement to fine-grained coordinate prediction. To address this problem, we construct CrossPoint-378K, a dataset with 378K question-answering pairs across 900 scenes, focused on actionable affordance regions that better reflect real-world manipulation and interaction scenarios. Furthermore, we propose CroPond that trained on the CrossPoint-378K dataset. Our CroPond achieves state-of-the-art performance on CrossPoint-Bench, surpassing Gemini-2.5-Pro by 39.7% accuracy, which offers a foundation for advancing future work on cross-view correspondence. The benchmark, dataset, and model are publicly available at https://github.com/WangYipu2002/CrossPoint.

cs.CV↗

arXiv2025

OpenCUA: Open Foundations for Computer-Use Agents

Vision-language models have demonstrated impressive capabilities as computer-use agents (CUAs) capable of automating diverse computer tasks. As their commercial potential grows, critical details of the most capable CUA systems remain closed. As these agents will increasingly mediate digital interactions and execute consequential decisions on our behalf, the research community needs access to open CUA frameworks to study their capabilities, limitations, and risks. To bridge this gap, we propose OpenCUA, a comprehensive open-source framework for scaling CUA data and foundation models. Our framework consists of: (1) an annotation infrastructure that seamlessly captures human computer-use demonstrations; (2) AgentNet, the first large-scale computer-use task dataset spanning 3 operating systems and 200+ applications and websites; (3) a scalable pipeline that transforms demonstrations into state-action pairs with reflective long Chain-of-Thought reasoning that sustain robust performance gains as data scales. Our end-to-end agent models demonstrate strong performance across CUA benchmarks. In particular, OpenCUA-72B achieves an average success rate of 45.0% on OSWorld-Verified, establishing a new state-of-the-art (SOTA) among open-source models. Further analysis confirms that our approach generalizes well across domains and benefits significantly from increased test-time computation. We release our annotation tool, datasets, code, and models to build open foundations for further CUA research.

cs.AI↗

arXiv2025

VisualTrans: A Benchmark for Real-World Visual Transformation Reasoning

Visual transformation reasoning (VTR) is a vital cognitive capability that empowers intelligent agents to understand dynamic scenes, model causal relationships, and predict future states, and thereby guiding actions and laying the foundation for advanced intelligent systems. However, existing benchmarks suffer from a sim-to-real gap, limited task complexity, and incomplete reasoning coverage, limiting their practical use in real-world scenarios. To address these limitations, we introduce VisualTrans, the first comprehensive benchmark specifically designed for VTR in real-world human-object interaction scenarios. VisualTrans encompasses 12 semantically diverse manipulation tasks and systematically evaluates three essential reasoning dimensions - spatial, procedural, and quantitative - through 6 well-defined subtask types. The benchmark features 472 high-quality question-answer pairs in various formats, including multiple-choice, open-ended counting, and target enumeration. We introduce a scalable data construction pipeline built upon first-person manipulation videos, which integrates task selection, image pair extraction, automated metadata annotation with large multimodal models, and structured question generation. Human verification ensures the final benchmark is both high-quality and interpretable. Evaluations of various state-of-the-art vision-language models show strong performance in static spatial tasks. However, they reveal notable shortcomings in dynamic, multi-step reasoning scenarios, particularly in areas like intermediate state recognition and transformation sequence planning. These findings highlight fundamental weaknesses in temporal modeling and causal reasoning, providing clear directions for future research aimed at developing more capable and generalizable VTR systems. The dataset and code are available at https://github.com/WangYipu2002/VisualTrans.

cs.CV↗

arXiv2024

Complexity Classification of Product State Problems for Local Hamiltonians

Product states, unentangled tensor products of single qubits, are a ubiquitous ansatz in quantum computation, including for state-of-the-art Hamiltonian approximation algorithms. A natural question is whether we should expect to efficiently solve product state problems on any interesting families of Hamiltonians. We completely classify the complexity of finding minimum-energy product states for Hamiltonians defined by any fixed set of allowed 2-qubit interactions. Our results follow a line of work classifying the complexity of solving Hamiltonian problems and classical constraint satisfaction problems based on the allowed constraints. We prove that estimating the minimum energy of a product state is in P if and only if all allowed interactions are 1-local, and NP-complete otherwise. Equivalently, any family of non-trivial two-body interactions generates Hamiltonians with NP-complete product-state problems. Our hardness constructions only require coupling strengths of constant magnitude. A crucial component of our proofs is a collection of hardness results for a new variant of the Vector Max-Cut problem, which should be of independent interest. Our definition involves sums of distances rather than squared distances and allows linear stretches. We similarly give a proof that the original Vector Max-Cut problem is NP-complete in 3 dimensions. This implies that optimizing over product states for Quantum Max-Cut (the quantum Heisenberg model) is NP-complete, even when every term is guaranteed to have positive unit weight.

quant-ph↗

arXiv2023

Tutorial: Nonlinear magnonics

Nonlinear magnonics studies the nonlinear interaction between magnons and other physical platforms (phonon, photon, qubit, spin texture) to generate novel magnon states for information processing. In this tutorial, we first introduce the nonlinear interactions of magnons in pure magnetic systems and hybrid magnon-phonon and magnon-photon systems. Then we show how these nonlinear interactions can generate exotic magnonic phenomena. In the classical regime, we will cover the parametric excitation of magnons, bistability and multistability, and the magnonic frequency comb. In the quantum regime, we will discuss the single magnon state, Schrödinger cat state and the entanglement and quantum steering among magnons, photons and phonons. The applications of the hybrid magnonics systems in quantum transducer and sensing will also be presented. Finally, we outlook the future development direction of nonlinear magnonics.

cond-mat.mes-hall↗

arXiv2022

Stochastic Neuromorphic Circuits for Solving MAXCUT

Finding the maximum cut of a graph (MAXCUT) is a classic optimization problem that has motivated parallel algorithm development. While approximate algorithms to MAXCUT offer attractive theoretical guarantees and demonstrate compelling empirical performance, such approximation approaches can shift the dominant computational cost to the stochastic sampling operations. Neuromorphic computing, which uses the organizing principles of the nervous system to inspire new parallel computing architectures, offers a possible solution. One ubiquitous feature of natural brains is stochasticity: the individual elements of biological neural networks possess an intrinsic randomness that serves as a resource enabling their unique computational capacities. By designing circuits and algorithms that make use of randomness similarly to natural brains, we hypothesize that the intrinsic randomness in microelectronics devices could be turned into a valuable component of a neuromorphic architecture enabling more efficient computations. Here, we present neuromorphic circuits that transform the stochastic behavior of a pool of random devices into useful correlations that drive stochastic solutions to MAXCUT. We show that these circuits perform favorably in comparison to software solvers and argue that this neuromorphic hardware implementation provides a path for scaling advantages. This work demonstrates the utility of combining neuromorphic principles with intrinsic randomness as a computational resource for new computational architectures.

cs.NE↗

arXiv2021

Maximum Integer Flows in Directed Planar Graphs with Multiple Sources and Sinks and Vertex Capacities

We consider the problem of finding maximum flows in planar graphs with capacities on both vertices and edges and with multiple sources and sinks. We present three algorithms when the capacities are integers. The first algorithm runs in $O(n \log^3 n + kn)$ time when all capacities are bounded, where $n$ is the number of vertices in the graph and $k$ is the number of terminals. This algorithm is the first to solve the vertex-disjoint paths problem in near-linear time when $k$ is bounded but larger than 2. The second algorithm runs in $O(k^2(k^3 + Δ) n \text{ polylog} (nU))$ time, where $U$ is the largest finite capacity of a single vertex and $Δ$ is the maximum degree of a vertex. Finally, when $k=3$, we present an algorithm that runs in $O(n \log n)$ time; this algorithm works even when the capacities are arbitrary reals. Our algorithms improve on the fastest previously known algorithms when $k$ and $Δ$ are small and $U$ is bounded by a polynomial in $n$. Prior to this result, the fastest algorithms ran in $O(n^2 / \log n)$ time for real capacities and $O(n^{3/2} \log n \log U)$ for integer capacities.

cs.DS↗

arXiv2019

The Single-Face Ideal Orientation Problem in Planar Graphs

We consider the ideal orientation problem in planar graphs. In this problem, we are given an undirected graph $G$ with positive edge lengths and $k$ pairs of distinct vertices $(s_1, t_1), \dots, (s_k, t_k)$ called terminals, and we want to assign an orientation to each edge such that for all $i$ the distance from $s_i$ to $t_i$ is preserved or report that no such orientation exists. We show that the problem is NP-hard in planar graphs. On the other hand, we show that the problem is polynomial-time solvable in planar graphs when $k$ is fixed, the vertices $s_1, t_1, \dots, s_k, t_k$ are all on the same face, and no two of terminal pairs cross (a pair $(s_i, t_i)$ crosses $(s_j, t_j)$ if the cyclic order of the vertices is $s_i,s_j,t_i,t_j$). For serial instances, we give a simpler and faster algorithm running in $O(n \log n)$ time, even if $k$ is part of the input. (An instance is serial if the terminals appear in cyclic order $u_1, v_1, \dots, u_k, v_k$, where for each $i$ we have either $(u_i, v_i) = (s_i, t_i)$ or $(u_i, v_i) = (t_i, s_i)$.) Finally, we consider a generalization of the problem in which the sum of the distances from $s_i$ to $t_i$ is to be minimized; in this case we give an algorithm for serial instances running in $O(kn^5)$ time.

cs.DS↗

arXiv2019

Topologically Trivial Closed Walks in Directed Surface Graphs

Let $G$ be a directed graph with $n$ vertices and $m$ edges, embedded on a surface $S$, possibly with boundary, with first Betti number $β$. We consider the complexity of finding closed directed walks in $G$ that are either contractible (trivial in homotopy) or bounding (trivial in integer homology) in $S$. Specifically, we describe algorithms to determine whether $G$ contains a simple contractible cycle in $O(n+m)$ time, or a contractible closed walk in $O(n+m)$ time, or a bounding closed walk in $O(β(n+m))$ time. Our algorithms rely on subtle relationships between strong connectivity in $G$ and in the dual graph $G^*$; our contractible-closed-walk algorithm also relies on a seminal topological result of Hass and Scott. We also prove that detecting simple bounding cycles is NP-hard. We also describe three polynomial-time algorithms to compute shortest contractible closed walks, depending on whether the fundamental group of the surface is free, abelian, or hyperbolic. A key step in our algorithm for hyperbolic surfaces is the construction of a context-free grammar with $O(g^2L^2)$ non-terminals that generates all contractible closed walks of length at most L, and only contractible closed walks, in a system of quads of genus $g\ge2$. Finally, we show that computing shortest simple contractible cycles, shortest simple bounding cycles, and shortest bounding closed walks are all NP-hard.

cs.CG↗