arXiv subjects

Cheng Peng

Publications and source records attributed to Cheng Peng.

At least 73 records · Page 4Linked to original sources

arXiv2025

SPIDER: Spatial Image CorresponDence Estimator for Robust Calibration

Reliable image correspondences form the foundation of vision-based spatial perception, enabling recovery of 3D structure and camera poses. However, unconstrained feature matching across domains such as aerial, indoor, and outdoor scenes remains challenging due to large variations in appearance, scale and viewpoint. Feature matching has been conventionally formulated as a 2D-to-2D problem; however, recent 3D foundation models provides spatial feature matching properties based on two-view geometry. While powerful, we observe that these spatially coherent matches often concentrate on dominant planar regions, e.g., walls or ground surfaces, while being less sensitive to fine-grained geometric details, particularly under large viewpoint changes. To better understand these trade-offs, we first perform linear probe experiments to evaluate the performance of various vision foundation models for image matching. Building on these insights, we introduce SPIDER, a universal feature matching framework that integrates a shared feature extraction backbone with two specialized network heads for estimating both 2D-based and 3D-based correspondences from coarse to fine. Finally, we introduce an image-matching evaluation benchmark that focuses on unconstrained scenarios with large baselines. SPIDER significantly outperforms SoTA methods, demonstrating its strong ability as a universal image-matching method.

cs.CV↗

arXiv2025

FlexAvatar: Flexible Large Reconstruction Model for Animatable Gaussian Head Avatars with Detailed Deformation

We present FlexAvatar, a flexible large reconstruction model for high-fidelity 3D head avatars with detailed dynamic deformation from single or sparse images, without requiring camera poses or expression labels. It leverages a transformer-based reconstruction model with structured head query tokens as canonical anchor to aggregate flexible input-number-agnostic, camera-pose-free and expression-free inputs into a robust canonical 3D representation. For detailed dynamic deformation, we introduce a lightweight UNet decoder conditioned on UV-space position maps, which can produce detailed expression-dependent deformations in real time. To better capture rare but critical expressions like wrinkles and bared teeth, we also adopt a data distribution adjustment strategy during training to balance the distribution of these expressions in the training set. Moreover, a lightweight 10-second refinement can further enhances identity-specific details in extreme identities without affecting deformation quality. Extensive experiments demonstrate that our FlexAvatar achieves superior 3D consistency, detailed dynamic realism compared with previous methods, providing a practical solution for animatable 3D avatar creation.

cs.CV↗

arXiv2025

Harnessing Negative Signals: Reinforcement Distillation from Teacher Data for LLM Reasoning

Recent advances in model distillation show that data from advanced reasoning models can effectively train smaller student models. However, standard practices discard incorrect reasoning traces -- valuable, yet underutilized data. This paper addresses the critical question: How can both positive and negative distilled reasoning traces be effectively leveraged to maximize LLM reasoning performance in an offline setting? We employ a two-stage training recipe: first, Supervised Fine-Tuning (SFT) on positive traces, followed by a refinement stage using both positive and negative traces. We find that a simple REINFORCE-style objective, which we term the Reinforcement Distillation (REDI) objective, outperforms established preference optimization methods like DPO and SimPO in this distillation context. Our empirical evaluations demonstrate the effectiveness of this approach. Notably, our Qwen-REDI-1.5B model, trained on just 131k traces from the open Open-R1 dataset, achieves an 83.1% score on MATH-500. Its performance matches that of DeepSeek-R1-Distill-Qwen-1.5B, a model trained on 800k proprietary data. This result showcases the remarkable data efficiency of utilizing previously discarded negative traces.

cs.LG↗

arXiv2025

Bi-Isolated d.c.e. Degrees and $Σ_1$ Induction

A Turing degree is d.c.e. if it contains a set that is the difference of two c.e. sets. A d.c.e. degree $\mathbf{d}$ is isolated if there exists a c.e. degree $\mathbf{a}<\mathbf{d}$ such that every c.e. degree below $\mathbf{d}$ is also below $\mathbf{a}$; $\mathbf{d}$ is upper isolated if there exists a c.e. degree $\mathbf{a}>\mathbf{d}$ such that every c.e. degree above $\mathbf{d}$ is also above $\mathbf{a}$; $\mathbf{d}$ is bi-isolated if it is both isolated and upper isolated. In this paper, we prove the existence of bi-isolated d.c.e. degrees in models of $\mathsf{I}Σ_1$.

math.LO↗

arXiv2025

D-commuting SYK model: building quantum chaos from integrable blocks

We construct a new family of quantum chaotic models by combining multiple copies of integrable commuting SYK models. As each copy of the commuting SYK model does not commute with others, this construction breaks the integrability of each commuting SYK and the family of models demonstrates the emergence of quantum chaos. We study the spectrum of this model analytically in the double-scaled limit. As the number of copies tends to infinity, the spectrum becomes compact and equivalent to the regular SYK model. For finite $d$ copies, the spectrum is close to the regular SYK model in UV but has an exponential tail $e^{E/T_c}$ in the IR. We identify the reciprocal of the exponent in the tail as a critical temperature $T_c$, above which the model should be quantum chaotic. $T_c$ monotonically decreases as $d$ increases, which expands the chaotic regime over the non-chaotic regime. We propose the existence of a new phase around $T_c$, and the dynamics should be very different in two phases. We further carry out numeric analysis at finite $d$, which supports our proposal. Given any finite dimensional local Hamiltonian, by decomposing it into $d$ groups, in which all terms in one group commute with each other but terms from different groups may not, our analysis can give an estimate of the critical temperature for quantum chaos based on the decomposition. We also comment on the implication of the critical temperature to future quantum simulations of quantum chaos and quantum gravity.

hep-th↗

arXiv2025

Scenes as Tokens: Multi-Scale Normal Distributions Transform Tokenizer for General 3D Vision-Language Understanding

Recent advances in 3D vision-language models (VLMs) highlight a strong potential for 3D scene understanding and reasoning. However, effectively tokenizing 3D scenes into holistic scene tokens, and leveraging these tokens across diverse 3D understanding tasks, remain highly challenging. We present NDTokenizer3D, a generalist 3D VLM that performs a wide range of 3D scene understanding tasks while naturally supporting human interactions, thereby bridging language-level reasoning with 3D spatial understanding. The core of our approach is a novel three-stage scene tokenization pipeline built upon a Multi-Scale Normal Distributions Transform (NDT) representation, paired with a Multi-Scale NDT Decoder (MSDec). Specifically, NDTokenizer3D first constructs a multi-scale NDT representation from raw high-resolution point clouds, preserving both global context and fine-grained geometric details. Next, the MSDec progressively fuses cross-scale NDT features, producing holistic scene tokens consumable by LLM endpoints. Beyond tokenization, MSDec is repurposed as a general interface for human-interactive prompting (points, boxes, masks) and segmentation-mask decoding, unifying diverse 3D scene understanding tasks within a single architecture. With this compact and unified design, NDTokenizer3D offers a fine-grained, general-purpose 3D VLM, achieving remarkable improvements in 3D Referring Segmentation, 3D Visual Question Answering, and 3D Dense Captioning.

cs.CV↗

arXiv2025

Comparative study and critical assessment of phase-field lattice Boltzmann models for laminar and turbulent two-phase flow simulations

Phase field lattice Boltzmann (LB) models have undergone continuous development, resulting in multiple variants widely used for simulating multiphase flows. However, direct performance comparisons remain limited, especially for three-dimensional cases. In this study, we present a systematic comparative analysis of several recent and representative phase-field LB models, covering four major categories: conservative Allen-Cahn, nonlocal Allen-Cahn, hybrid Allen-Cahn, and Cahn-Hilliard models. Their accuracy, numerical stability and mass/volume conservation are assessed through a series of canonical two-phase flow problems. Beyond the commonly tested two-dimensional laminar cases, we extend the evaluation to three-dimensional droplet-laden turbulent flows, which expose more critical limitations of the existing models. The results show that while all models perform satisfactorily in two dimensions, they still suffer from substantial droplet volume loss in turbulence, particularly at high Weber numbers. Overall, conservative Allen-Cahn-based LB models exhibit the most favorable balance of numerical stability, accuracy and computational efficiency.

physics.flu-dyn↗

arXiv2025

A Structure-Agnostic Co-Tuning Framework for LLMs and SLMs in Cloud-Edge Systems

The surge in intelligent applications driven by large language models (LLMs) has made it increasingly difficult for bandwidth-limited cloud servers to process extensive LLM workloads in real time without compromising user data privacy. To solve these problems, recent research has focused on constructing cloud-edge consortia that integrate server-based LLM with small language models (SLMs) on mobile edge devices. Furthermore, designing collaborative training mechanisms within such consortia to enhance inference performance has emerged as a promising research direction. However, the cross-domain deployment of SLMs, coupled with structural heterogeneity in SLMs architectures, poses significant challenges to enhancing model performance. To this end, we propose Co-PLMs, a novel co-tuning framework for collaborative training of large and small language models, which integrates the process of structure-agnostic mutual learning to realize knowledge exchange between the heterogeneous language models. This framework employs distilled proxy models (DPMs) as bridges to enable collaborative training between the heterogeneous server-based LLM and on-device SLMs, while preserving the domain-specific insights of each device. The experimental results show that Co-PLMs outperform state-of-the-art methods, achieving average increases of 5.38% in Rouge-L and 4.88% in EM.

cs.DC↗

arXiv2025

FusionRF: High-Fidelity Satellite Neural Radiance Fields from Multispectral and Panchromatic Acquisitions

We introduce FusionRF, a novel framework for digital surface reconstruction from satellite multispectral and panchromatic images. Current work has demonstrated the increased accuracy of neural photogrammetry for surface reconstruction from optical satellite images compared to algorithmic methods. Common satellites produce both a panchromatic and multispectral image, which contain high spatial and spectral information respectively. Current neural reconstruction methods require multispectral images to be upsampled with a pansharpening method using the spatial data in the panchromatic image. However, these methods may introduce biases and hallucinations due to domain gaps. FusionRF introduces joint image fusion during optimization through a novel cross-resolution kernel that learns to resolve spatial resolution loss present in multispectral images. As input, FusionRF accepts the original multispectral and panchromatic data, eliminating the need for image preprocessing. FusionRF also leverages multimodal appearance embeddings that encode the image characteristics of each modality and view within a uniform representation. By optimizing on both modalities, FusionRF learns to fuse image modalities while performing reconstruction tasks and eliminates the need for a pansharpening preprocessing step. We evaluate our method on multispectral and panchromatic satellite images from the WorldView-3 satellite in various locations, and show that FusionRF provides an average of 17% reduction in depth reconstruction error, and renders sharp training and novel views.

cs.CV↗

arXiv2025

RoboOS-NeXT: A Unified Memory-based Framework for Lifelong, Scalable, and Robust Multi-Robot Collaboration

The proliferation of collaborative robots across diverse tasks and embodiments presents a central challenge: achieving lifelong adaptability, scalable coordination, and robust scheduling in multi-agent systems. Existing approaches, from vision-language-action (VLA) models to hierarchical frameworks, fall short due to their reliance on limited or dividual-agent memory. This fundamentally constrains their ability to learn over long horizons, scale to heterogeneous teams, or recover from failures, highlighting the need for a unified memory representation. To address these limitations, we introduce RoboOS-NeXT, a unified memory-based framework for lifelong, scalable, and robust multi-robot collaboration. At the core of RoboOS-NeXT is the novel Spatio-Temporal-Embodiment Memory (STEM), which integrates spatial scene geometry, temporal event history, and embodiment profiles into a shared representation. This memory-centric design is integrated into a brain-cerebellum framework, where a high-level brain model performs global planning by retrieving and updating STEM, while low-level controllers execute actions locally. This closed loop between cognition, memory, and execution enables dynamic task allocation, fault-tolerant collaboration, and consistent state synchronization. We conduct extensive experiments spanning complex coordination tasks in restaurants, supermarkets, and households. Our results demonstrate that RoboOS-NeXT achieves superior performance across heterogeneous embodiments, validating its effectiveness in enabling lifelong, scalable, and robust multi-robot collaboration. Project website: https://flagopen.github.io/RoboOS/

cs.RO↗

arXiv2025

MS-GS: Multi-Appearance Sparse-View 3D Gaussian Splatting in the Wild

In-the-wild photo collections often contain limited volumes of imagery and exhibit multiple appearances, e.g., taken at different times of day or seasons, posing significant challenges to scene reconstruction and novel view synthesis. Although recent adaptations of Neural Radiance Field (NeRF) and 3D Gaussian Splatting (3DGS) have improved in these areas, they tend to oversmooth and are prone to overfitting. In this paper, we present MS-GS, a novel framework designed with Multi-appearance capabilities in Sparse-view scenarios using 3DGS. To address the lack of support due to sparse initializations, our approach is built on the geometric priors elicited from monocular depth estimations. The key lies in extracting and utilizing local semantic regions with a Structure-from-Motion (SfM) points anchored algorithm for reliable alignment and geometry cues. Then, to introduce multi-view constraints, we propose a series of geometry-guided supervision steps at virtual views in pixel and feature levels to encourage 3D consistency and reduce overfitting. We also introduce a dataset and an in-the-wild experiment setting to set up more realistic benchmarks. We demonstrate that MS-GS achieves photorealistic renderings under various challenging sparse-view and multi-appearance conditions, and outperforms existing approaches significantly across different datasets.

cs.CV↗

arXiv2025

Aria Gen 2 Pilot Dataset

The Aria Gen 2 Pilot Dataset (A2PD) is an egocentric multimodal open dataset captured using the state-of-the-art Aria Gen 2 glasses. To facilitate timely access, A2PD is released incrementally with ongoing dataset enhancements. The initial release features Dia'ane, our primary subject, who records her daily activities alongside friends, each equipped with Aria Gen 2 glasses. It encompasses five primary scenarios: cleaning, cooking, eating, playing, and outdoor walking. In each of the scenarios, we provide comprehensive raw sensor data and output data from various machine perception algorithms. These data illustrate the device's ability to perceive the wearer, the surrounding environment, and interactions between the wearer and the environment, while maintaining robust performance across diverse users and conditions. The A2PD is publicly available at projectaria.com, with open-source tools and usage examples provided in Project Aria Tools.

cs.CV↗

arXiv2025

Gauging the complex SYK model

Motivated by SYK-like models describing near-BPS black holes in string/M-theory, we consider gauging the U$(1)$ symmetry of the complex SYK model in the presence of a Wilson line with charge $k$. At a fixed background gauge field, solutions to the Schwinger-Dyson equations display vastly different properties from those at a fixed real chemical potential. In the partition function and the two-point function, the integral over the gauge field is performed either directly or via a large $N$ saddle point approximation, and both results are consistent with exact diagonalization data. From the behaviour of the two-point function at large $N$, we deduce that the conformal symmetry at low energies is preserved at fixed $κ= k/N = 0$, but broken at $κ\neq 0$. In addition, we find that there is maximal chaos for all $k$.

hep-th↗

arXiv2025

EvoWorld: Evolving Panoramic World Generation with Explicit 3D Memory

Humans possess a remarkable ability to mentally explore and replay 3D environments they have previously experienced. Inspired by this mental process, we present EvoWorld: a world model that bridges panoramic video generation with evolving 3D memory to enable spatially consistent long-horizon exploration. Given a single panoramic image as input, EvoWorld first generates future video frames by leveraging a video generator with fine-grained view control, then evolves the scene's 3D reconstruction using a feedforward plug-and-play transformer, and finally synthesizes futures by conditioning on geometric reprojections from this evolving explicit 3D memory. Unlike prior state-of-the-arts that synthesize videos only, our key insight lies in exploiting this evolving 3D reconstruction as explicit spatial guidance for the video generation process, projecting the reconstructed geometry onto target viewpoints to provide rich spatial cues that significantly enhance both visual realism and geometric consistency. To evaluate long-range exploration capabilities, we introduce the first comprehensive benchmark spanning synthetic outdoor environments, Habitat indoor scenes, and challenging real-world scenarios, with particular emphasis on loop-closure detection and spatial coherence over extended trajectories. Extensive experiments demonstrate that our evolving 3D memory substantially improves visual fidelity and maintains spatial scene coherence compared to existing approaches, representing a significant advance toward long-horizon spatially consistent world modeling.

cs.CV↗

arXiv2025

A Study of Large Language Models for Patient Information Extraction: Model Architecture, Fine-Tuning Strategy, and Multi-task Instruction Tuning

Natural language processing (NLP) is a key technology to extract important patient information from clinical narratives to support healthcare applications. The rapid development of large language models (LLMs) has revolutionized many NLP tasks in the clinical domain, yet their optimal use in patient information extraction tasks requires further exploration. This study examines LLMs' effectiveness in patient information extraction, focusing on LLM architectures, fine-tuning strategies, and multi-task instruction tuning techniques for developing robust and generalizable patient information extraction systems. This study aims to explore key concepts of using LLMs for clinical concept and relation extraction tasks, including: (1) encoder-only or decoder-only LLMs, (2) prompt-based parameter-efficient fine-tuning (PEFT) algorithms, and (3) multi-task instruction tuning on few-shot learning performance. We benchmarked a suite of LLMs, including encoder-based LLMs (BERT, GatorTron) and decoder-based LLMs (GatorTronGPT, Llama 3.1, GatorTronLlama), across five datasets. We compared traditional full-size fine-tuning and prompt-based PEFT. We explored a multi-task instruction tuning framework that combines both tasks across four datasets to evaluate the zero-shot and few-shot learning performance using the leave-one-dataset-out strategy.

cs.CL↗

arXiv2025

Computational approaches for virus host prediction: A review of methods and applications

Accurate prediction of virus-host interactions is critical for understanding viral ecology and developing applications like phage therapy. However, the growing number of computational tools has created a complex landscape, making direct performance comparison challenging due to inconsistent benchmarks and varying usability. Here, we provide a systematic review and a rigorous benchmark of 27 virus-host prediction tools. We formulate the host prediction task into two primary frameworks, link prediction and multi-class classification, and construct two benchmark datasets to evaluate tool performance in distinct scenarios: a database-centric dataset (RefSeq-VHDB) and a metagenomic discovery dataset (MetaHiC-VHDB). Our results reveal that no single tool is universally optimal. Performance is highly context-dependent, with tools like CHERRY and iPHoP demonstrating robust, broad applicability, while others, such as RaFAH and PHIST, excel in specific contexts. We further identify a critical trade-off between predictive accuracy, prediction rate, and computational cost. This work serves as a practical guide for researchers and establishes a standardized benchmark to drive future innovation in deciphering complex virus-host interactions.

q-bio.GN↗

arXiv2025

Mechanism Design for Congested Facility Location

This paper investigates mechanism design for congested facility location problems, where agents are partitioned into groups with conflicting interests (e.g., competition for booking a basketball court in a gymnasium), and each agent's cost increases when the facility is located closer to their competitors. We analyze three types of misreporting: location-only, group-membership-only, and combined misreporting. To minimize social cost, we propose a strategyproof mechanism that achieves optimality under location-only misreporting. For group membership and combined misreporting, we show that the median mechanism attains tight approximation bounds. For the objective of minimizing maximum cost, we introduce novel strategyproof mechanisms for location-only and group-membership-only misreporting, while employing the leftmost mechanism under combined misreporting. We prove that all proposed mechanisms achieve nearly tight performance guarantees.

cs.GT↗

arXiv2025

Isolated d.c.e. degrees and $Σ_1$ induction

A Turing degree is d.c.e. if it contains a set that is the difference of two c.e. sets. A d.c.e. degree $\mathbf{d}$ is isolated by a c.e. degree $\mathbf{a}<\mathbf{d}$ if all c.e. degrees that are below $\mathbf{d}$ are also below $\mathbf{a}$; $\mathbf{d}$ is isolated from above by a c.e. degree $\mathbf{a}>\mathbf{d}$ if all c.e. degrees that are above $\mathbf{d}$ are also above $\mathbf{a}$. In this paper, we study the inductive strength of both isolated and upper isolated d.c.e. degrees from the point of view of reverse recursion theory. We show that (1) $P^{-} + BΣ_1 + \text{Exp} \vdash IΣ_1 \leftrightarrow$ There is an isolated proper d.c.e. degree below $\mathbf{0}'$; (2) $P^{-} + BΣ_1 + \text{Exp} \vdash IΣ_1 \leftrightarrow$ There is an upper isolated proper d.c.e. degree below $\mathbf{0}'$.

math.LO↗