arXiv subjects

Lingyu Kong

Publications and source records attributed to Lingyu Kong.

At least 19 recordsLinked to original sources

arXiv2026

MoGe-3: Fine-Detail Monocular Geometry Estimation with Self-Guided Sparse Volumetric Refinement

Monocular geometry estimation has recently achieved impressive performance across diverse scenes. However, state-of-the-art models still face notable distortion in local 3D structure, especially in fine details, like thin structures and small objects. We attribute this limitation to an architectural mismatch: most current models decode 3D geometry within a 2D parameterization, where feature interactions are governed by image-plane proximity rather than true 3D spatial relationships. This inadvertently mixes features from geometrically distant surfaces, resulting in over-smoothed geometry particularly around thin or elongated structure. In this paper, we propose MoGe-3, a fine-detail monocular geometry estimation model with Self-Guided Sparse 3D Refinement (SSR) that lifts monocular geometry modeling from 2D image space to 3D space for high-fidelity metric-scale point maps. MoGe-3 lifts the coarse point map from a foundation base model onto a sparse voxel shell and refines it via SSR. The SSR employs sparse convolutions that aggregate features based on 3D spatial locality, avoiding feature mixing across depth discontinuities. Extensive experiments on diverse datasets demonstrate that MoGe-3 significantly outperforms existing approaches in recovering fine detailed 3D geometry across both quantitative metrics and qualitative visualizations. Project page: https://qft-333.github.io/moge3page/

cs.CV↗

arXiv2026

IDEAL: In-DEpth ALignment Makes A Discrete Representation AutoEncoder

Built on pretrained vision foundation models (VFMs), representation autoencoders (RAEs) have recently emerged as a promising approach for constructing semantically rich latent spaces for image generation. However, their reconstruction quality often remains suboptimal, largely because deep VFM representations do not preserve sufficient fine-grained visual detail. This limitation becomes even more severe after discretization, where missing low-level information is difficult to recover. In fact, we observe that shallow VFM features retain considerably richer local appearance and structural detail, which complements the high-level semantics carried by deep features used in existing RAEs. Motivated by this complementary property, we propose Ideal, an In-depth Alignment framework for discrete representation autoencoding. By jointly aligning quantized tokens with both shallow and deep VFM features, Ideal enables the resulting discrete visual tokens to preserve both visual fidelity and rich semantics. Extensive experiments demonstrate that Ideal yields superior reconstruction performance, achieving 0.61 rFID on ImageNet and outperforming the previous best method by 0.28. When used for autoregressive image generation, Ideal further produces a gFID of 1.89, establishing a new state of the art for autoregressive image generation.

cs.CV↗

arXiv2026

MatterSim-MT: A multi-task foundation model for in silico materials characterization

Accurate property characterization is a major bottleneck in materials design. While first-principles methods and task-specific machine-learning models have driven important progress, they remain fundamentally limited in scalability and generalizability across the vast space of structures and properties relevant to real-world materials design. We present MatterSim-MT, a multi-task foundation model for in silico materials simulation and property characterization. The model is pretrained on over 35 million first-principles-labeled structures covering 89 elements, temperatures up to 5000 K and pressures up to 1000 GPa, and is fine-tuned on various properties including Bader charges, magnetic moments, Born effective charges, and dielectric matrices. Out of the box, MatterSim-MT not only serves as a foundation model for predicting material structure, dynamics and thermodynamics, its multi-task architecture also enables a wide range of complex simulations that cannot be captured by potential energy surfaces alone. For example, we demonstrate pressure-dependent LO-TO phonon splitting in SiC with close agreement with experiment, electric hysteresis in ferroelectric BaTiO3, and the cationic-to-anionic redox transition during delithiation of a Li-rich cathode material. Finally, we show that MatterSim-MT scales well with more data and parameters, can be efficiently fine-tuned to higher levels of theory, and can be efficiently extended to new systems via active learning. Overall, we believe this approach provides a scalable route to accurate in silico materials characterization.

cond-mat.mtrl-sci↗

arXiv2025

Polymer-inspired mechanical metamaterials

Metamaterials benefit from unique architected patterns to achieve lightweight with exceptional mechanical properties inaccessible to conventional materials. Typical mechanical metamaterials are inspired by crystal-like lattice structures, whose closely packed frameworks often exhibit a rigid mechanical nature. Here, we present polymer-inspired metamaterials (PIMs) by programming deformation and strengthening mechanisms that mimic the mechanical roles of key constituent elements in polymer networks. By combining metamaterial programmability with polymer-inspired structures, we design crosslinking, proto-crystalline order, and entanglement in PIMs to enable macroscale strengthening mechanisms inspired by crosslink, molecular-density, and pre-stretch strengthening in polymers, expanding the metamaterial structure-property design space. This macroscale polymer-inspired programmability also suggests that PIMs could serve as a design platform incorporating the programmability strategies to achieve desired deformation and strengthening responses, holding a potential for applications in soft robotic joints and compliant connectors.

physics.app-ph↗

arXiv2025

Scalable Foundation Interatomic Potentials via Message-Passing Pruning and Graph Partitioning

Atomistic foundation models (AFMs) have great promise as accurate interatomic potentials, and have enabled data-efficient molecular dynamics simulations with near quantum mechanical accuracy. However, AFMs remain markedly slower at inference and are far more memory-intensive than conventional interatomic potentials, due to the need to capture a wide range of chemical and structural motifs in pre-training datasets requiring deep, parameter-rich model architectures. These deficiencies currently limit the practical use of AFMs in molecular dynamics (MD) simulations at extended temporal and spatial scales. To address this problem, we propose a general workflow for accelerating and scaling AFMs containing message-passing architectures. We find that removing low-contribution message-passing layers from AFM backbones serves as an effective pruning method, significantly reducing the parameter count while preserving the accuracy and data-efficiency of AFMs. Once pruned, these models become more accessible for large scale simulations via a graph-partitioned, GPU-distributed strategy, which we implement and demonstrate within the AFM fine-tuning platform MatterTune. We show that this approach supports million-atom simulations on both single and multiple GPUs, and enables task-specific large-scale simulations at nanosecond timescales with AFM-level accuracy.

cond-mat.mtrl-sci↗

arXiv2025

OccVLA: Vision-Language-Action Model with Implicit 3D Occupancy Supervision

Multimodal large language models (MLLMs) have shown strong vision-language reasoning abilities but still lack robust 3D spatial understanding, which is critical for autonomous driving. This limitation stems from two key challenges: (1) the difficulty of constructing accessible yet effective 3D representations without expensive manual annotations, and (2) the loss of fine-grained spatial details in VLMs due to the absence of large-scale 3D vision-language pretraining. To address these challenges, we propose OccVLA, a novel framework that integrates 3D occupancy representations into a unified multimodal reasoning process. Unlike prior approaches that rely on explicit 3D inputs, OccVLA treats dense 3D occupancy as both a predictive output and a supervisory signal, enabling the model to learn fine-grained spatial structures directly from 2D visual inputs. The occupancy predictions are regarded as implicit reasoning processes and can be skipped during inference without performance degradation, thereby adding no extra computational overhead. OccVLA achieves state-of-the-art results on the nuScenes benchmark for trajectory planning and demonstrates superior performance on 3D visual question-answering tasks, offering a scalable, interpretable, and fully vision-based solution for autonomous driving.

cs.AI↗

arXiv2025

A Comprehensive Assessment and Benchmark Study of Large Atomistic Foundation Models for Phonons

The rapid development of universal machine learning potentials (uMLPs) has enabled efficient, accurate predictions of diverse material properties across broad chemical spaces. While their capability for modeling phonon properties is emerging, systematic benchmarking across chemically diverse systems remains limited. We evaluate six recent uMLPs (EquiformerV2, MatterSim, MACE, and CHGNet) on 2,429 crystalline materials from the Open Quantum Materials Database. Models were used to compute atomic forces in displaced supercells, derive interatomic force constants (IFCs), and predict phonon properties including lattice thermal conductivity (LTC), compared with density functional theory (DFT) and experimental data. The EquiformerV2 pretrained model trained on the OMat24 dataset exhibits strong performance in predicting atomic forces and third-order IFC, while its fine-tuned counterpart consistently outperforms other models in predicting second-order IFC, LTC, and other phonon properties. Although MACE and CHGNet demonstrated comparable force prediction accuracy to EquiformerV2, notable discrepancies in IFC fitting led to poor LTC predictions. Conversely, MatterSim, despite lower force accuracy, achieved intermediate IFC predictions, suggesting error cancellation and complex relationships between force accuracy and phonon predictions. This benchmark guides the evaluation and selection of uMLPs for high-throughput screening of materials with targeted thermal transport properties.

cond-mat.mtrl-sci↗

arXiv2025

Clapper: Compact Learning and Video Representation in VLMs

Current vision-language models (VLMs) have demonstrated remarkable capabilities across diverse video understanding applications. Designing VLMs for video inputs requires effectively modeling the temporal dimension (i.e. capturing dependencies across frames) and balancing the processing of short and long videos. Specifically, short videos demand preservation of fine-grained details, whereas long videos require strategic compression of visual information to handle extensive temporal contexts efficiently. However, our empirical analysis reveals a critical limitation: most existing VLMs suffer severe performance degradation in long video understanding tasks when compressing visual tokens below a quarter of their original visual tokens. To enable more effective modeling of both short and long video inputs, we propose Clapper, a method that utilizes a slow-fast strategy for video representation and introduces a novel module named TimePerceiver for efficient temporal-spatial encoding within existing VLM backbones. By using our method, we achieves 13x compression of visual tokens per frame (averaging 61 tokens/frame) without compromising QA accuracy. In our experiments, Clapper achieves 62.0% on VideoMME, 69.8% on MLVU, and 67.4% on TempCompass, all with fewer than 6,000 visual tokens per video. The code will be publicly available on the homepage.

cs.CV↗

arXiv2025

MatterTune: An Integrated, User-Friendly Platform for Fine-Tuning Atomistic Foundation Models to Accelerate Materials Simulation and Discovery

Geometric machine learning models such as graph neural networks have achieved remarkable success in recent years in chemical and materials science research for applications such as high-throughput virtual screening and atomistic simulations. The success of these models can be attributed to their ability to effectively learn latent representations of atomic structures directly from the training data. Conversely, this also results in high data requirements for these models, hindering their application to problems which are data sparse which are common in this domain. To address this limitation, there is a growing development in the area of pre-trained machine learning models which have learned general, fundamental, geometric relationships in atomistic data, and which can then be fine-tuned to much smaller application-specific datasets. In particular, models which are pre-trained on diverse, large-scale atomistic datasets have shown impressive generalizability and flexibility to downstream applications, and are increasingly referred to as atomistic foundation models. To leverage the untapped potential of these foundation models, we introduce MatterTune, a modular and extensible framework that provides advanced fine-tuning capabilities and seamless integration of atomistic foundation models into downstream materials informatics and simulation workflows, thereby lowering the barriers to adoption and facilitating diverse applications in materials science. In its current state, MatterTune supports a number of state-of-the-art foundation models such as ORB, MatterSim, JMP, and EquformerV2, and hosts a wide range of features including a modular and flexible design, distributed and customizable fine-tuning, broad support for downstream informatics tasks, and more.

cond-mat.mtrl-sci↗

arXiv2025

Pix2Cap-COCO: Advancing Visual Comprehension via Pixel-Level Captioning

We present Pix2Cap-COCO, the first panoptic pixel-level caption dataset designed to advance fine-grained visual understanding. To achieve this, we carefully design an automated annotation pipeline that prompts GPT-4V to generate pixel-aligned, instance-specific captions for individual objects within images, enabling models to learn more granular relationships between objects and their contexts. This approach results in 167,254 detailed captions, with an average of 22.94 words per caption. Building on Pix2Cap-COCO, we introduce a novel task, panoptic segmentation-captioning, which challenges models to recognize instances in an image and provide detailed descriptions for each simultaneously. To benchmark this task, we design a robust baseline based on X-Decoder. The experimental results demonstrate that Pix2Cap-COCO is a particularly challenging dataset, as it requires models to excel in both fine-grained visual understanding and detailed language generation. Furthermore, we leverage Pix2Cap-COCO for Supervised Fine-Tuning (SFT) on large multimodal models (LMMs) to enhance their performance. For example, training with Pix2Cap-COCO significantly improves the performance of GPT4RoI, yielding gains in CIDEr +1.4%, ROUGE +0.4%, and SPICE +0.5% on Visual Genome dataset, and strengthens its region understanding ability on the ViP-BENCH, with an overall improvement of +5.1%, including notable increases in recognition accuracy +11.2% and language generation quality +22.2%.

cs.CV↗

arXiv2025

FOCUS: Towards Universal Foreground Segmentation

Foreground segmentation is a fundamental task in computer vision, encompassing various subdivision tasks. Previous research has typically designed task-specific architectures for each task, leading to a lack of unification. Moreover, they primarily focus on recognizing foreground objects without effectively distinguishing them from the background. In this paper, we emphasize the importance of the background and its relationship with the foreground. We introduce FOCUS, the Foreground ObjeCts Universal Segmentation framework that can handle multiple foreground tasks. We develop a multi-scale semantic network using the edge information of objects to enhance image features. To achieve boundary-aware segmentation, we propose a novel distillation method, integrating the contrastive learning strategy to refine the prediction mask in multi-modal feature space. We conduct extensive experiments on a total of 13 datasets across 5 tasks, and the results demonstrate that FOCUS consistently outperforms the state-of-the-art task-specific models on most metrics.

cs.CV↗

arXiv2024

General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model

Traditional OCR systems (OCR-1.0) are increasingly unable to meet people's usage due to the growing demand for intelligent processing of man-made optical characters. In this paper, we collectively refer to all artificial optical signals (e.g., plain texts, math/molecular formulas, tables, charts, sheet music, and even geometric shapes) as "characters" and propose the General OCR Theory along with an excellent model, namely GOT, to promote the arrival of OCR-2.0. The GOT, with 580M parameters, is a unified, elegant, and end-to-end model, consisting of a high-compression encoder and a long-contexts decoder. As an OCR-2.0 model, GOT can handle all the above "characters" under various OCR tasks. On the input side, the model supports commonly used scene- and document-style images in slice and whole-page styles. On the output side, GOT can generate plain or formatted results (markdown/tikz/smiles/kern) via an easy prompt. Besides, the model enjoys interactive OCR features, i.e., region-level recognition guided by coordinates or colors. Furthermore, we also adapt dynamic resolution and multi-page OCR technologies to GOT for better practicality. In experiments, we provide sufficient results to prove the superiority of our model.

cs.CV↗

arXiv2024

Expectation value estimation with parametrized quantum circuits

Estimating properties of quantum states, such as fidelities, molecular energies, and correlation functions, is a fundamental task in quantum information science. Due to the limitation of practical quantum devices, including limited circuit depth and connectivity, estimating even linear properties encounters high sample complexity. To address this inefficiency, we propose a framework that optimizes sample complexity for estimating the expectation value of any observable using a shallow parameterized quantum circuit. Within this framework, we introduce two decomposition algorithms, a tensor network approach and a greedy projection approach that decompose the target observable into a linear combination of multiple observables, each of which can be diagonalized with the shallow circuit. Using this decomposition, we then apply an importance sampling algorithm to estimate the expectation value of the target observable. We numerically demonstrate the performance of our algorithm by estimating the expectation values of some specific Hamiltonians and inner product of a Slater determinant with a pure state, highlighting advantages compared to some conventional methods. Additionally, we derive the fundamental lower bound for the sample complexity required to estimate a target observable using a given shallow quantum circuit, thereby enhancing our understanding of the capabilities of shallow circuits in quantum learning tasks.

quant-ph↗

arXiv2024

Merlin:Empowering Multimodal LLMs with Foresight Minds

Humans possess the remarkable ability to foresee the future to a certain extent based on present observations, a skill we term as foresight minds. However, this capability remains largely under explored within existing Multimodal Large Language Models (MLLMs), hindering their capacity to learn the fundamental principles of how things operate and the intentions behind the observed subjects. To address this issue, we introduce the integration of future modeling into the existing learning frameworks of MLLMs. By utilizing the subject trajectory, a highly structured representation of a consecutive frame sequence, as a learning objective, we aim to bridge the gap between the past and the future. We propose two innovative methods to empower MLLMs with foresight minds, Foresight Pre-Training (FPT) and Foresight Instruction-Tuning (FIT), which are inspired by the modern learning paradigm of LLMs. Specifically, FPT jointly training various tasks centered on trajectories, enabling MLLMs to learn how to attend and predict entire trajectories from a given initial observation. Then, FIT requires MLLMs to first predict trajectories of related objects and then reason about potential future events based on them. Aided by FPT and FIT, we build a novel and unified MLLM named Merlin that supports multi-images input and analysis about potential actions of multiple objects for the future reasoning. Experimental results show Merlin powerful foresight minds with impressive performance on both future reasoning and visual comprehension tasks.

cs.CV↗

arXiv2024

Focus Anywhere for Fine-grained Multi-page Document Understanding

Modern LVLMs still struggle to achieve fine-grained document understanding, such as OCR/translation/caption for regions of interest to the user, tasks that require the context of the entire page, or even multiple pages. Accordingly, this paper proposes Fox, an effective pipeline, hybrid data, and tuning strategy, that catalyzes LVLMs to focus anywhere on single/multi-page documents. We introduce a novel task to boost the document understanding by making LVLMs focus attention on the document-level region, such as redefining full-page OCR as foreground focus. We employ multiple vision vocabularies to extract visual hybrid knowledge for interleaved document pages (e.g., a page containing a photo). Meanwhile, we render cross-vocabulary vision data as the catalyzer to achieve a full reaction of multiple visual vocabularies and in-document figure understanding. Further, without modifying the weights of multiple vision vocabularies, the above catalyzed fine-grained understanding capabilities can be efficiently tuned to multi-page documents, enabling the model to focus anywhere in both format-free and page-free manners. Besides, we build a benchmark including 9 fine-grained sub-tasks (e.g., region-level OCR/summary, color-guided OCR) to promote document analysis in the community. The experimental results verify the superiority of our model.

cs.CV↗

arXiv2024

MatterSim: A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures

Accurate and fast prediction of materials properties is central to the digital transformation of materials design. However, the vast design space and diverse operating conditions pose significant challenges for accurately modeling arbitrary material candidates and forecasting their properties. We present MatterSim, a deep learning model actively learned from large-scale first-principles computations, for efficient atomistic simulations at first-principles level and accurate prediction of broad material properties across the periodic table, spanning temperatures from 0 to 5000 K and pressures up to 1000 GPa. Out-of-the-box, the model serves as a machine learning force field, and shows remarkable capabilities not only in predicting ground-state material structures and energetics, but also in simulating their behavior under realistic temperatures and pressures, signifying an up to ten-fold enhancement in precision compared to the prior best-in-class. This enables MatterSim to compute materials' lattice dynamics, mechanical and thermodynamic properties, and beyond, to an accuracy comparable with first-principles methods. Specifically, MatterSim predicts Gibbs free energies for a wide range of inorganic solids with near-first-principles accuracy and achieves a 15 meV/atom resolution for temperatures up to 1000K compared with experiments. This opens an opportunity to predict experimental phase diagrams of materials at minimal computational cost. Moreover, MatterSim also serves as a platform for continuous learning and customization by integrating domain-specific data. The model can be fine-tuned for atomistic simulations at a desired level of theory or for direct structure-to-property predictions, achieving high data efficiency with a reduction in data requirements by up to 97%.

cond-mat.mtrl-sci↗

arXiv2024

OneChart: Purify the Chart Structural Extraction via One Auxiliary Token

Chart parsing poses a significant challenge due to the diversity of styles, values, texts, and so forth. Even advanced large vision-language models (LVLMs) with billions of parameters struggle to handle such tasks satisfactorily. To address this, we propose OneChart: a reliable agent specifically devised for the structural extraction of chart information. Similar to popular LVLMs, OneChart incorporates an autoregressive main body. Uniquely, to enhance the reliability of the numerical parts of the output, we introduce an auxiliary token placed at the beginning of the total tokens along with an additional decoder. The numerically optimized (auxiliary) token allows subsequent tokens for chart parsing to capture enhanced numerical features through causal attention. Furthermore, with the aid of the auxiliary token, we have devised a self-evaluation mechanism that enables the model to gauge the reliability of its chart parsing results by providing confidence scores for the generated content. Compared to current state-of-the-art (SOTA) chart parsing models, e.g., DePlot, ChartVLM, ChartAst, OneChart significantly outperforms in Average Precision (AP) for chart structural extraction across multiple public benchmarks, despite enjoying only 0.2 billion parameters. Moreover, as a chart parsing agent, it also brings 10%+ accuracy gains for the popular LVLM (LLaVA-1.6) in the downstream ChartQA benchmark.

cs.CV↗

arXiv2024

Small Language Model Meets with Reinforced Vision Vocabulary

Playing Large Vision Language Models (LVLMs) in 2023 is trendy among the AI community. However, the relatively large number of parameters (more than 7B) of popular LVLMs makes it difficult to train and deploy on consumer GPUs, discouraging many researchers with limited resources. Imagine how cool it would be to experience all the features of current LVLMs on an old GTX1080ti (our only game card). Accordingly, we present Vary-toy in this report, a small-size Vary along with Qwen-1.8B as the base ``large'' language model. In Vary-toy, we introduce an improved vision vocabulary, allowing the model to not only possess all features of Vary but also gather more generality. Specifically, we replace negative samples of natural images with positive sample data driven by object detection in the procedure of generating vision vocabulary, more sufficiently utilizing the capacity of the vocabulary network and enabling it to efficiently encode visual information corresponding to natural objects. For experiments, Vary-toy can achieve 65.6% ANLS on DocVQA, 59.1% accuracy on ChartQA, 88.1% accuracy on RefCOCO, and 29% on MMVet. The code will be publicly available on the homepage.

cs.CV↗