arXiv subjects

Xinlei Chen

Publications and source records attributed to Xinlei Chen.

At least 91 records · Page 5Linked to original sources

arXiv2025

How to Enable LLM with 3D Capacity? A Survey of Spatial Reasoning in LLM

3D spatial understanding is essential in real-world applications such as robotics, autonomous vehicles, virtual reality, and medical imaging. Recently, Large Language Models (LLMs), having demonstrated remarkable success across various domains, have been leveraged to enhance 3D understanding tasks, showing potential to surpass traditional computer vision methods. In this survey, we present a comprehensive review of methods integrating LLMs with 3D spatial understanding. We propose a taxonomy that categorizes existing methods into three branches: image-based methods deriving 3D understanding from 2D visual data, point cloud-based methods working directly with 3D representations, and hybrid modality-based methods combining multiple data streams. We systematically review representative methods along these categories, covering data representations, architectural modifications, and training strategies that bridge textual and 3D modalities. Finally, we discuss current limitations, including dataset scarcity and computational challenges, while highlighting promising research directions in spatial perception, multi-modal fusion, and real-world applications.

cs.CV↗

arXiv2025

Scaling Language-Free Visual Representation Learning

Visual Self-Supervised Learning (SSL) currently underperforms Contrastive Language-Image Pretraining (CLIP) in multimodal settings such as Visual Question Answering (VQA). This multimodal gap is often attributed to the semantics introduced by language supervision, even though visual SSL and CLIP models are often trained on different data. In this work, we ask the question: "Do visual self-supervised approaches lag behind CLIP due to the lack of language supervision, or differences in the training data?" We study this question by training both visual SSL and CLIP models on the same MetaCLIP data, and leveraging VQA as a diverse testbed for vision encoders. In this controlled setup, visual SSL models scale better than CLIP models in terms of data and model capacity, and visual SSL performance does not saturate even after scaling up to 7B parameters. Consequently, we observe visual SSL methods achieve CLIP-level performance on a wide range of VQA and classic vision benchmarks. These findings demonstrate that pure visual SSL can match language-supervised visual pretraining at scale, opening new opportunities for vision-centric representation learning.

cs.CV↗

arXiv2025

MR-COGraphs: Communication-efficient Multi-Robot Open-vocabulary Mapping System via 3D Scene Graphs

Collaborative perception in unknown environments is crucial for multi-robot systems. With the emergence of foundation models, robots can now not only perceive geometric information but also achieve open-vocabulary scene understanding. However, existing map representations that support open-vocabulary queries often involve large data volumes, which becomes a bottleneck for multi-robot transmission in communication-limited environments. To address this challenge, we develop a method to construct a graph-structured 3D representation called COGraph, where nodes represent objects with semantic features and edges capture their spatial adjacency relationships. Before transmission, a data-driven feature encoder is applied to compress the feature dimensions of the COGraph. Upon receiving COGraphs from other robots, the semantic features of each node are recovered using a decoder. We also propose a feature-based approach for place recognition and translation estimation, enabling the merging of local COGraphs into a unified global map. We validate our framework on two realistic datasets and the real-world environment. The results demonstrate that, compared to existing baselines for open-vocabulary map construction, our framework reduces the data volume by over 80\% while maintaining mapping and query performance without compromise. For more details, please visit our website at https://github.com/efc-robot/MR-COGraphs.

cs.RO↗

arXiv2025

An Image is Worth More Than 16x16 Patches: Exploring Transformers on Individual Pixels

This work does not introduce a new method. Instead, we present an interesting finding that questions the necessity of the inductive bias of locality in modern computer vision architectures. Concretely, we find that vanilla Transformers can operate by directly treating each individual pixel as a token and achieve highly performant results. This is substantially different from the popular design in Vision Transformer, which maintains the inductive bias from ConvNets towards local neighborhoods (e.g. by treating each 16x16 patch as a token). We showcase the effectiveness of pixels-as-tokens across three well-studied computer vision tasks: supervised learning for classification and regression, self-supervised learning via masked autoencoding, and image generation with diffusion models. Although it's computationally less practical to directly operate on individual pixels, we believe the community must be made aware of this surprising piece of knowledge when devising the next generation of neural network architectures for computer vision.

cs.CV↗

arXiv2025

Underwater Motions Analysis and Control of a Coupling-Tiltable Unmanned Aerial-Aquatic Vehicle

Coupling-Tiltable Unmanned Aerial-Aquatic Vehicles (UAAVs) have gained increasing importance, yet lack comprehensive analysis and suitable controllers. This paper analyzes the underwater motion characteristics of a self-designed UAAV, Mirs-Alioth, and designs a controller for it. The effectiveness of the controller is validated through experiments. The singularities of Mirs-Alioth are derived as Singular Thrust Tilt Angle (STTA), which serve as an essential tool for an analysis of its underwater motion characteristics. The analysis reveals several key factors for designing the controller. These include the need for logic switching, using a Nussbaum function to compensate control direction uncertainty in the auxiliary channel, and employing an auxiliary controller to mitigate coupling effects. Based on these key points, a control scheme is designed. It consists of a controller that regulates the thrust tilt angle to the singular value, an auxiliary controller incorporating a Saturated Nussbaum function, and a logic switch. Eventually, two sets of experiments are conducted to validate the effectiveness of the controller and demonstrate the necessity of the Nussbaum function.

cs.RO↗

arXiv2025

UrbanVideo-Bench: Benchmarking Vision-Language Models on Embodied Intelligence with Video Data in Urban Spaces

Large multimodal models exhibit remarkable intelligence, yet their embodied cognitive abilities during motion in open-ended urban 3D space remain to be explored. We introduce a benchmark to evaluate whether video-large language models (Video-LLMs) can naturally process continuous first-person visual observations like humans, enabling recall, perception, reasoning, and navigation. We have manually control drones to collect 3D embodied motion video data from real-world cities and simulated environments, resulting in 1.5k video clips. Then we design a pipeline to generate 5.2k multiple-choice questions. Evaluations of 17 widely-used Video-LLMs reveal current limitations in urban embodied cognition. Correlation analysis provides insight into the relationships between different tasks, showing that causal reasoning has a strong correlation with recall, perception, and navigation, while the abilities for counterfactual and associative reasoning exhibit lower correlation with other tasks. We also validate the potential for Sim-to-Real transfer in urban embodiment through fine-tuning.

cs.CV↗

arXiv2025

SniffySquad: Patchiness-Aware Gas Source Localization with Multi-Robot Collaboration

Gas source localization is pivotal for the rapid mitigation of gas leakage disasters, where mobile robots emerge as a promising solution. However, existing methods predominantly schedule robots' movements based on reactive stimuli or simplified gas plume models. These approaches typically excel in idealized, simulated environments but fall short in real-world gas environments characterized by their patchy distribution. In this work, we introduce SniffySquad, a multi-robot olfaction-based system designed to address the inherent patchiness in gas source localization. SniffySquad incorporates a patchiness-aware active sensing approach that enhances the quality of data collection and estimation. Moreover, it features an innovative collaborative role adaptation strategy to boost the efficiency of source-seeking endeavors. Extensive evaluations demonstrate that our system achieves an increase in the success rate by $20\%+$ and an improvement in path efficiency by $30\%+$, outperforming state-of-the-art gas source localization solutions.

cs.RO↗

arXiv2025

PLPHP: Per-Layer Per-Head Vision Token Pruning for Efficient Large Vision-Language Models

Large Vision-Language Models (LVLMs) have demonstrated remarkable capabilities across a range of multimodal tasks. However, their inference efficiency is constrained by the large number of visual tokens processed during decoding. To address this challenge, we propose Per-Layer Per-Head Vision Token Pruning (PLPHP), a two-level fine-grained pruning method including Layer-Level Retention Rate Allocation and Head-Level Vision Token Pruning. Motivated by the Vision Token Re-attention phenomenon across decoder layers, we dynamically adjust token retention rates layer by layer. Layers that exhibit stronger attention to visual information preserve more vision tokens, while layers with lower vision attention are aggressively pruned. Furthermore, PLPHP applies pruning at the attention head level, enabling different heads within the same layer to independently retain critical context. Experiments on multiple benchmarks demonstrate that PLPHP delivers an 18% faster decoding speed and reduces the Key-Value Cache (KV Cache) size by over 50%, all at the cost of 0.46% average performance drop, while also achieving notable performance improvements in multi-image tasks. These results highlight the effectiveness of fine-grained token pruning and contribute to advancing the efficiency and scalability of LVLMs. Our source code will be made publicly available.

cs.CV↗

arXiv2025

Ultra-High-Frequency Harmony: mmWave Radar and Event Camera Orchestrate Accurate Drone Landing

For precise, efficient, and safe drone landings, ground platforms should real-time, accurately locate descending drones and guide them to designated spots. While mmWave sensing combined with cameras improves localization accuracy, the lower sampling frequency of traditional frame cameras compared to mmWave radar creates bottlenecks in system throughput. In this work, we replace the traditional frame camera with event camera, a novel sensor that harmonizes in sampling frequency with mmWave radar within the ground platform setup, and introduce mmE-Loc, a high-precision, low-latency ground localization system designed for drone landings. To fully leverage the \textit{temporal consistency} and \textit{spatial complementarity} between these modalities, we propose two innovative modules, \textit{consistency-instructed collaborative tracking} and \textit{graph-informed adaptive joint optimization}, for accurate drone measurement extraction and efficient sensor fusion. Extensive real-world experiments in landing scenarios from a leading drone delivery company demonstrate that mmE-Loc outperforms state-of-the-art methods in both localization accuracy and latency.

cs.RO↗

arXiv2025

Understanding and Evaluating Hallucinations in 3D Visual Language Models

Recently, 3D-LLMs, which combine point-cloud encoders with large models, have been proposed to tackle complex tasks in embodied intelligence and scene understanding. In addition to showing promising results on 3D tasks, we found that they are significantly affected by hallucinations. For instance, they may generate objects that do not exist in the scene or produce incorrect relationships between objects. To investigate this issue, this work presents the first systematic study of hallucinations in 3D-LLMs. We begin by quickly evaluating hallucinations in several representative 3D-LLMs and reveal that they are all significantly affected by hallucinations. We then define hallucinations in 3D scenes and, through a detailed analysis of datasets, uncover the underlying causes of these hallucinations. We find three main causes: (1) Uneven frequency distribution of objects in the dataset. (2) Strong correlations between objects. (3) Limited diversity in object attributes. Additionally, we propose new evaluation metrics for hallucinations, including Random Point Cloud Pair and Opposite Question Evaluations, to assess whether the model generates responses based on visual information and aligns it with the text's meaning.

cs.CV↗

arXiv2025

Learning Real-World Action-Video Dynamics with Heterogeneous Masked Autoregression

We propose Heterogeneous Masked Autoregression (HMA) for modeling action-video dynamics to generate high-quality data and evaluation in scaling robot learning. Building interactive video world models and policies for robotics is difficult due to the challenge of handling diverse settings while maintaining computational efficiency to run in real time. HMA uses heterogeneous pre-training from observations and action sequences across different robotic embodiments, domains, and tasks. HMA uses masked autoregression to generate quantized or soft tokens for video predictions. \ourshort achieves better visual fidelity and controllability than the previous robotic video generation models with 15 times faster speed in the real world. After post-training, this model can be used as a video simulator from low-level action inputs for evaluating policies and generating synthetic data. See this link https://liruiw.github.io/hma for more information.

cs.RO↗

arXiv2025

LLMs can see and hear without any training

We present MILS: Multimodal Iterative LLM Solver, a surprisingly simple, training-free approach, to imbue multimodal capabilities into your favorite LLM. Leveraging their innate ability to perform multi-step reasoning, MILS prompts the LLM to generate candidate outputs, each of which are scored and fed back iteratively, eventually generating a solution to the task. This enables various applications that typically require training specialized models on task-specific data. In particular, we establish a new state-of-the-art on emergent zero-shot image, video and audio captioning. MILS seamlessly applies to media generation as well, discovering prompt rewrites to improve text-to-image generation, and even edit prompts for style transfer! Finally, being a gradient-free optimization approach, MILS can invert multimodal embeddings into text, enabling applications like cross-modal arithmetic.

cs.CV↗

arXiv2025

Learnings from Scaling Visual Tokenizers for Reconstruction and Generation

Visual tokenization via auto-encoding empowers state-of-the-art image and video generative models by compressing pixels into a latent space. Although scaling Transformer-based generators has been central to recent advances, the tokenizer component itself is rarely scaled, leaving open questions about how auto-encoder design choices influence both its objective of reconstruction and downstream generative performance. Our work aims to conduct an exploration of scaling in auto-encoders to fill in this blank. To facilitate this exploration, we replace the typical convolutional backbone with an enhanced Vision Transformer architecture for Tokenization (ViTok). We train ViTok on large-scale image and video datasets far exceeding ImageNet-1K, removing data constraints on tokenizer scaling. We first study how scaling the auto-encoder bottleneck affects both reconstruction and generation -- and find that while it is highly correlated with reconstruction, its relationship with generation is more complex. We next explored the effect of separately scaling the auto-encoders' encoder and decoder on reconstruction and generation performance. Crucially, we find that scaling the encoder yields minimal gains for either reconstruction or generation, while scaling the decoder boosts reconstruction but the benefits for generation are mixed. Building on our exploration, we design ViTok as a lightweight auto-encoder that achieves competitive performance with state-of-the-art auto-encoders on ImageNet-1K and COCO reconstruction tasks (256p and 512p) while outperforming existing auto-encoders on 16-frame 128p video reconstruction for UCF-101, all with 2-5x fewer FLOPs. When integrated with Diffusion Transformers, ViTok demonstrates competitive performance on image generation for ImageNet-1K and sets new state-of-the-art benchmarks for class-conditional video generation on UCF-101.

cs.CV↗

arXiv2025

Gaussian Masked Autoencoders

This paper explores Masked Autoencoders (MAE) with Gaussian Splatting. While reconstructive self-supervised learning frameworks such as MAE learns good semantic abstractions, it is not trained for explicit spatial awareness. Our approach, named Gaussian Masked Autoencoder, or GMAE, aims to learn semantic abstractions and spatial understanding jointly. Like MAE, it reconstructs the image end-to-end in the pixel space, but beyond MAE, it also introduces an intermediate, 3D Gaussian-based representation and renders images via splatting. We show that GMAE can enable various zero-shot learning capabilities of spatial understanding (e.g., figure-ground segmentation, image layering, edge detection, etc.) while preserving the high-level semantics of self-supervised representation quality from MAE. To our knowledge, we are the first to employ Gaussian primitives in an image representation learning framework beyond optimization-based single-scene reconstructions. We believe GMAE will inspire further research in this direction and contribute to developing next-generation techniques for modeling high-fidelity visual data. More details at https://brjathu.github.io/gmae

cs.CV↗

arXiv2025

Test-Time Training on Video Streams

Prior work has established Test-Time Training (TTT) as a general framework to further improve a trained model at test time. Before making a prediction on each test instance, the model is first trained on the same instance using a self-supervised task such as reconstruction. We extend TTT to the streaming setting, where multiple test instances - video frames in our case - arrive in temporal order. Our extension is online TTT: The current model is initialized from the previous model, then trained on the current frame and a small window of frames immediately before. Online TTT significantly outperforms the fixed-model baseline for four tasks, on three real-world datasets. The improvements are more than 2.2x and 1.5x for instance and panoptic segmentation. Surprisingly, online TTT also outperforms its offline variant that accesses strictly more information, training on all frames from the entire test video regardless of temporal order. This finding challenges those in prior work using synthetic videos. We formalize a notion of locality as the advantage of online over offline TTT, and analyze its role with ablations and a theory based on bias-variance trade-off.

cs.CV↗

arXiv2024

MetaMorph: Multimodal Understanding and Generation via Instruction Tuning

In this work, we propose Visual-Predictive Instruction Tuning (VPiT) - a simple and effective extension to visual instruction tuning that enables a pretrained LLM to quickly morph into an unified autoregressive model capable of generating both text and visual tokens. VPiT teaches an LLM to predict discrete text tokens and continuous visual tokens from any input sequence of image and text data curated in an instruction-following format. Our empirical investigation reveals several intriguing properties of VPiT: (1) visual generation ability emerges as a natural byproduct of improved visual understanding, and can be unlocked efficiently with a small amount of generation data; (2) while we find understanding and generation to be mutually beneficial, understanding data contributes to both capabilities more effectively than generation data. Building upon these findings, we train our MetaMorph model and achieve competitive performance on both visual understanding and generation. In visual generation, MetaMorph can leverage the world knowledge and reasoning abilities gained from LLM pretraining, and overcome common failure modes exhibited by other generation models. Our results suggest that LLMs may have strong "prior" vision capabilities that can be efficiently adapted to both visual understanding and generation with a relatively simple instruction tuning process.

cs.CV↗

arXiv2024

Simultaneous Multiband Photometry of the Early Optical Afterglow of GRB 240825A with Mephisto

Gamma-ray bursts (GRBs) are the most luminous transients in the universe. The interaction of the relativistic jet with the circumburst medium produces an afterglow and generates multiwavelength emission. In this work, we present simultaneous multiband photometry of GRB~240825A with the Multi-channel Photometric Survey Telescope (Mephisto) and analyze its temporal and spectral properties. The measurement began 128 seconds after the GRB trigger and continued until the fourth day when the afterglow essentially diminished and the measured brightness was close to that of the host galaxy. Based on the multiband light curves in the $uvgriz$ bands, we find that the optical flux density satisfies $F_{ν,{\rm obs}}\propto t^{-1.34}ν^{-2.48}$ with a spectral index of $2.48$ much larger than those of most other GRBs. To reconcile the measured much softer spectral energy distribution (SED) with that predicted by the standard afterglow model, an extra host-galaxy extinction of $E_{B-V}\sim(0.37-0.57)$ mag is required. We interpreted this excess as arising from a dense circumburst medium. We further find that the SED of the optical afterglow hardened as the afterglow decayed and the color excess $E_{B-V}$ decreased $\sim0.26$ mag from 100 seconds to 3000 seconds after the GRB trigger. Finally, we analyze the properties of the host galaxy of GRB~240825A based on data from the SDSS, PanSTARRS and HSC-SSP surveys. For a host redshift of $z=0.659$, the stellar mass and star formation rate of the host galaxy are estimated to be $\log(M_*/M_\odot)=10.0^{+0.3}_{-0.3}$ and $\log({\rm SFR}/M_{\odot}{\rm yr}^{-1})= 0.6^{+0.8}_{-3.3}$, respectively, pointing to a gas-rich, star-forming, medium-size galaxy.

astro-ph.HE↗

arXiv2024

Learning Video Representations without Natural Videos

We show that useful video representations can be learned from synthetic videos and natural images, without incorporating natural videos in the training. We propose a progression of video datasets synthesized by simple generative processes, that model a growing set of natural video properties (e.g., motion, acceleration, and shape transformations). The downstream performance of video models pre-trained on these generated datasets gradually increases with the dataset progression. A VideoMAE model pre-trained on our synthetic videos closes 97.2\% of the performance gap on UCF101 action classification between training from scratch and self-supervised pre-training from natural videos, and outperforms the pre-trained model on HMDB51. Introducing crops of static images to the pre-training stage results in similar performance to UCF101 pre-training and outperforms the UCF101 pre-trained model on 11 out of 14 out-of-distribution datasets of UCF101-P. Analyzing the low-level properties of the datasets, we identify correlations between frame diversity, frame similarity to natural data, and downstream performance. Our approach provides a more controllable and transparent alternative to video data curation processes for pre-training.

cs.CV↗