arXiv subjects

Tianyi Zhang

Publications and source records attributed to Tianyi Zhang.

At least 217 records · Page 12Linked to original sources

arXiv2024

Context-Aware Token Selection and Packing for Enhanced Vision Transformer

In recent years, the long-range attention mechanism of vision transformers has driven significant performance breakthroughs across various computer vision tasks. However, the traditional self-attention mechanism, which processes both informative and non-informative tokens, suffers from inefficiency and inaccuracies. While sparse attention mechanisms have been introduced to mitigate these issues by pruning tokens involved in attention, they often lack context-awareness and intelligence. These mechanisms frequently apply a uniform token selection strategy across different inputs for batch training or optimize efficiency only for the inference stage. To overcome these challenges, we propose a novel algorithm: Select and Pack Attention (SPA). SPA dynamically selects informative tokens using a low-cost gating layer supervised by selection labels and packs these tokens into new batches, enabling a variable number of tokens to be used in parallelized GPU batch training and inference. Extensive experiments across diverse datasets and computer vision tasks demonstrate that SPA delivers superior performance and efficiency, including a 0.6 mAP improvement in object detection and a 16.4% reduction in computational costs.

cs.CV↗

arXiv2024

AraSync: Precision Time Synchronization in Rural Wireless Living Lab

Time synchronization is a critical component in network operation and management, and it is also required by Ultra-Reliable, Low-Latency Communications (URLLC) in next-generation wireless systems such as those of 5G, 6G, and Open RAN. In this context, we design and implement AraSync as an end-to-end time synchronization system in the ARA wireless living lab to enable advanced wireless experiments and applications involving stringent time constraints. We make use of Precision Time Protocol (PTP) at different levels to achieve synchronization accuracy in the order of nanoseconds. Along with fiber networks, AraSync enables time synchronization across the AraHaul wireless x-haul network consisting of long-range, high-capacity mmWave and microwave links. In this paper, we present the detailed design and implementation of AraSync, including its hardware and software components and the PTP network topology. Further, we experimentally characterize the performance of AraSync from spatial and temporal dimensions. Our measurement and analysis of the clock offset and mean path delay show the impact of the wireless channel and weather conditions on the PTP synchronization accuracy.

cs.NI↗

arXiv2024

Nearly Optimal Dynamic Set Cover: Breaking the Quadratic-in-$f$ Time Barrier

The dynamic set cover problem has been subject to extensive research since the pioneering works of [Bhattacharya et al, 2015] and [Gupta et al, 2017]. The input is a set system $(U, S)$ on a fixed collection $S$ of sets and a dynamic universe of elements, where each element appears in a most $f$ sets and the cost of each set lies in the range $[1/C, 1]$, and the goal is to efficiently maintain an approximately-minimum set cover under insertions and deletions of elements. Most previous work considers the low-frequency regime, namely $f = O(\log n)$, and this line of work has culminated with a deterministic $(1+ε)f$-approximation algorithm with amortized update time $O(\frac{f^2}{ε^3} + \frac{f}{ε^2}\log C)$ [Bhattacharya et al, 2021]. In the high-frequency regime of $f = Ω(\log n)$, an $O(\log n)$-approximation algorithm with amortized update time $O(f\log n)$ was given by [Gupta et al, 2017]. Interestingly, at the intersection of the two regimes, i.e., $f = Θ(\log n)$, the state-of-the-art results coincide: approximation $Θ(f) = Θ(\log n)$ with amortized update time $O(f^2) = O(f \log n) = O(\log^2 n)$. Up to this date, no previous work achieved update time of $o(f^2)$. In this paper we break the $Ω(f^2)$ update time barrier via the following results: (1) $(1+ε)f$-approximation can be maintained in $O\left(\frac{f}{ε^3}\log^*f + \frac{f}{ε^3}\log C\right) = O_{ε,C}(f \log^* f)$ expected amortized update time; our algorithm works against an adaptive adversary. (2) $(1+ε)f$-approximation can be maintained deterministically in $O\left(\frac{1}εf\log f + \frac{f}{ε^3} + \frac{f\log C}{ε^2}\right) = O_{ε,C}(f \log f)$ amortized update time.

cs.DS↗

arXiv2024

A new approach for fine-tuning sentence transformers for intent classification and out-of-scope detection tasks

In virtual assistant (VA) systems it is important to reject or redirect user queries that fall outside the scope of the system. One of the most accurate approaches for out-of-scope (OOS) rejection is to combine it with the task of intent classification on in-scope queries, and to use methods based on the similarity of embeddings produced by transformer-based sentence encoders. Typically, such encoders are fine-tuned for the intent-classification task, using cross-entropy loss. Recent work has shown that while this produces suitable embeddings for the intent-classification task, it also tends to disperse in-scope embeddings over the full sentence embedding space. This causes the in-scope embeddings to potentially overlap with OOS embeddings, thereby making OOS rejection difficult. This is compounded when OOS data is unknown. To mitigate this issue our work proposes to regularize the cross-entropy loss with an in-scope embedding reconstruction loss learned using an auto-encoder. Our method achieves a 1-4% improvement in the area under the precision-recall curve for rejecting out-of-sample (OOS) instances, without compromising intent classification performance.

cs.CL↗

arXiv2024

Almost-Optimal Sublinear Additive Spanners

Given an undirected unweighted graph $G = (V, E)$ on $n$ vertices and $m$ edges, a subgraph $H\subseteq G$ is a spanner of $G$ with stretch function $f: \mathbb{R}_+ \rightarrow \mathbb{R}_+$, if for every pair $s, t$ of vertices in $V$, $\text{dist}_{H}(s, t)\le f(\text{dist}_{G}(s, t))$. When $f(d) = d + o(d)$, $H$ is called a sublinear additive spanner; when $f(d) = d + o(n)$, $H$ is called an \emph{additive spanner}, and $f(d) - d$ is usually called the \emph{additive stretch} of $H$. As our primary result, we show that for any constant $δ>0$ and constant integer $k\geq 2$, every graph on $n$ vertices has a sublinear additive spanner with stretch function $f(d)=d+O(d^{1-1/k})$ and $O\big(n^{1+\frac{1+δ}{2^{k+1}-1}}\big)$ edges. When $k = 2$, this improves upon the previous spanner construction with stretch function $f(d) = d + O(d^{1/2})$ and $\tilde{O}(n^{1+3/17})$ edges; for any constant integer $k\geq 3$, this improves upon the previous spanner construction with stretch function $f(d) = d + O(d^{1-1/k})$ and $O\bigg(n^{1+\frac{(3/4)^{k-2}}{7 - 2\cdot (3/4)^{k-2}}}\bigg)$ edges. Most importantly, the size of our spanners almost matches the lower bound of $Ω\big(n^{1+\frac{1}{2^{k+1}-1}}\big)$, which holds for all compression schemes achieving the same stretch function. As our second result, we show a new construction of additive spanners with stretch $O(n^{0.403})$ and $\tilde{O}(n)$ edges, which slightly improves upon the previous stretch bound of $O(n^{3/7+\varepsilon})$ achieved by linear-size spanners. An additional advantage of our spanner is that it admits a subquadratic construction runtime of $\tilde{O}(m + n^{13/7})$, while the previous construction requires all-pairs shortest paths computation which takes $O(\min\{mn, n^{2.373}\})$ time.

cs.DS↗

arXiv2024

Even Faster $(Δ+ 1)$-Edge Coloring via Shorter Multi-Step Vizing Chains

Vizing's Theorem from 1964 states that any $n$-vertex $m$-edge graph with maximum degree $Δ$ can be {\em edge colored} using at most $Δ+ 1$ colors. For over 40 years, the state-of-the-art running time for computing such a coloring, obtained independently by Arjomandi [1982] and by Gabow, Nishizeki, Kariv, Leven and Terada~[1985], was $\tilde O(m\sqrt{n})$. Very recently, this time bound was improved in two independent works, by Bhattacharya, Carmon, Costa, Solomon and Zhang to $\tilde O(mn^{1/3})$, and by Assadi to $\tilde O(n^2)$. In this paper we present an algorithm that computes such a coloring in $\tilde O(mn^{1/4})$ time. Our key technical contribution is a subroutine for extending the coloring to one more edge within time $\tilde O(Δ^2 + \sqrt{Δn})$. The best previous time bound of any color extension subroutine is either the trivial $O(n)$, dominated by the length of a Vizing chain, or the bound $\tilde{O}(Δ^6)$ by Bernshteyn [2022], dominated by the length of {\em multi-step Vizing chains}, which is basically a concatenation of multiple (carefully chosen) Vizing chains. Our color extension subroutine produces significantly shorter multi-step Vizing chains than in previous works, for sufficiently large $Δ$.

cs.DS↗

arXiv2024

Collu-Bench: A Benchmark for Predicting Language Model Hallucinations in Code

Despite their success, large language models (LLMs) face the critical challenge of hallucinations, generating plausible but incorrect content. While much research has focused on hallucinations in multiple modalities including images and natural language text, less attention has been given to hallucinations in source code, which leads to incorrect and vulnerable code that causes significant financial loss. To pave the way for research in LLMs' hallucinations in code, we introduce Collu-Bench, a benchmark for predicting code hallucinations of LLMs across code generation (CG) and automated program repair (APR) tasks. Collu-Bench includes 13,234 code hallucination instances collected from five datasets and 11 diverse LLMs, ranging from open-source models to commercial ones. To better understand and predict code hallucinations, Collu-Bench provides detailed features such as the per-step log probabilities of LLMs' output, token types, and the execution feedback of LLMs' generated code for in-depth analysis. In addition, we conduct experiments to predict hallucination on Collu-Bench, using both traditional machine learning techniques and neural networks, which achieves 22.03 -- 33.15% accuracy. Our experiments draw insightful findings of code hallucination patterns, reveal the challenge of accurately localizing LLMs' hallucinations, and highlight the need for more sophisticated techniques.

cs.SE↗

arXiv2024

Learning Shared RGB-D Fields: Unified Self-supervised Pre-training for Label-efficient LiDAR-Camera 3D Perception

Constructing large-scale labeled datasets for multi-modal perception model training in autonomous driving presents significant challenges. This has motivated the development of self-supervised pretraining strategies. However, existing pretraining methods mainly employ distinct approaches for each modality. In contrast, we focus on LiDAR-Camera 3D perception models and introduce a unified pretraining strategy, NeRF-Supervised Masked Auto Encoder (NS-MAE), which optimizes all modalities through a shared formulation. NS-MAE leverages NeRF's ability to encode both appearance and geometry, enabling efficient masked reconstruction of multi-modal data. Specifically, embeddings are extracted from corrupted LiDAR point clouds and images, conditioned on view directions and locations. Then, these embeddings are rendered into multi-modal feature maps from two crucial viewpoints for 3D driving perception: perspective and bird's-eye views. The original uncorrupted data serve as reconstruction targets for self-supervised learning. Extensive experiments demonstrate the superior transferability of NS-MAE across various 3D perception tasks under different fine-tuning settings. Notably, NS-MAE outperforms prior SOTA pre-training methods that employ separate strategies for each modality in BEV map segmentation under the label-efficient fine-tuning setting. Our code is publicly available at https://github.com/Xiaohao-Xu/Unified-Pretrain-AD/ .

cs.CV↗

arXiv2024

SplaTraj: Camera Trajectory Generation with Semantic Gaussian Splatting

Many recent developments for robots to represent environments have focused on photorealistic reconstructions. This paper particularly focuses on generating sequences of images from the photorealistic Gaussian Splatting models, that match instructions that are given by user-inputted language. We contribute a novel framework, SplaTraj, which formulates the generation of images within photorealistic environment representations as a continuous-time trajectory optimization problem. Costs are designed so that a camera following the trajectory poses will smoothly traverse through the environment and render the specified spatial information in a photogenic manner. This is achieved by querying a photorealistic representation with language embedding to isolate regions that correspond to the user-specified inputs. These regions are then projected to the camera's view as it moves over time and a cost is constructed. We can then apply gradient-based optimization and differentiate through the rendering to optimize the trajectory for the defined cost. The resulting trajectory moves to photogenically view each of the specified objects. We empirically evaluate our approach on a suite of environments and instructions, and demonstrate the quality of generated image sequences.

cs.RO↗

arXiv2024

PhotoReg: Photometrically Registering 3D Gaussian Splatting Models

Building accurate representations of the environment is critical for intelligent robots to make decisions during deployment. Advances in photorealistic environment models have enabled robots to develop hyper-realistic reconstructions, which can be used to generate images that are intuitive for human inspection. In particular, the recently introduced \ac{3DGS}, which describes the scene with up to millions of primitive ellipsoids, can be rendered in real time. \ac{3DGS} has rapidly gained prominence. However, a critical unsolved problem persists: how can we fuse multiple \ac{3DGS} into a single coherent model? Solving this problem will enable robot teams to jointly build \ac{3DGS} models of their surroundings. A key insight of this work is to leverage the {duality} between photorealistic reconstructions, which render realistic 2D images from 3D structure, and \emph{3D foundation models}, which predict 3D structure from image pairs. To this end, we develop PhotoReg, a framework to register multiple photorealistic \ac{3DGS} models with 3D foundation models. As \ac{3DGS} models are generally built from monocular camera images, they have \emph{arbitrary scale}. To resolve this, PhotoReg actively enforces scale consistency among the different \ac{3DGS} models by considering depth estimates within these models. Then, the alignment is iteratively refined with fine-grained photometric losses to produce high-quality fused \ac{3DGS} models. We rigorously evaluate PhotoReg on both standard benchmark datasets and our custom-collected datasets, including with two quadruped robots. The code is released at \url{ziweny11.github.io/photoreg}.

cs.RO↗

arXiv2024

PharmacyGPT: The AI Pharmacist

In this study, we introduce PharmacyGPT, a novel framework to assess the capabilities of large language models (LLMs) such as ChatGPT and GPT-4 in emulating the role of clinical pharmacists. Our methodology encompasses the utilization of LLMs to generate comprehensible patient clusters, formulate medication plans, and forecast patient outcomes. We conduct our investigation using real data acquired from the intensive care unit (ICU) at the University of North Carolina Chapel Hill (UNC) Hospital. Our analysis offers valuable insights into the potential applications and limitations of LLMs in the field of clinical pharmacy, with implications for both patient care and the development of future AI-driven healthcare solutions. By evaluating the performance of PharmacyGPT, we aim to contribute to the ongoing discourse surrounding the integration of artificial intelligence in healthcare settings, ultimately promoting the responsible and efficacious use of such technologies.

cs.CL↗

arXiv2024

CodeJudge: Evaluating Code Generation with Large Language Models

Large Language Models (LLMs) have shown promising performance in code generation. However, how to reliably evaluate code generated by LLMs remains an unresolved problem. This paper presents CodeJudge, a code evaluation framework that leverages LLMs to evaluate the semantic correctness of generated code without the need for test cases. We investigate different ways to guide the LLM in performing "slow thinking" to arrive at an in-depth and reliable evaluation. We experimented with four LLMs as evaluators on four code generation datasets and five programming languages. The results show that CodeJudge significantly outperformed existing methods in most settings. Furthermore, compared with a SOTA GPT-3.5-based code evaluation method, CodeJudge achieved better results even when using a much smaller model, Llama-3-8B-Instruct. Our code and datasets are available on GitHub https://github.com/VichyTong/CodeJudge.

cs.LG↗

arXiv2024

Reasoning about the Unseen for Efficient Outdoor Object Navigation

Robots should exist anywhere humans do: indoors, outdoors, and even unmapped environments. In contrast, the focus of recent advancements in Object Goal Navigation(OGN) has targeted navigating in indoor environments by leveraging spatial and semantic cues that do not generalize outdoors. While these contributions provide valuable insights into indoor scenarios, the broader spectrum of real-world robotic applications often extends to outdoor settings. As we transition to the vast and complex terrains of outdoor environments, new challenges emerge. Unlike the structured layouts found indoors, outdoor environments lack clear spatial delineations and are riddled with inherent semantic ambiguities. Despite this, humans navigate with ease because we can reason about the unseen. We introduce a new task OUTDOOR, a new mechanism for Large Language Models (LLMs) to accurately hallucinate possible futures, and a new computationally aware success metric for pushing research forward in this more complex domain. Additionally, we show impressive results on both a simulated drone and physical quadruped in outdoor environments. Our agent has no premapping and our formalism outperforms naive LLM-based approaches

cs.RO↗

arXiv2024

Toward General-Purpose Robots via Foundation Models: A Survey and Meta-Analysis

Building general-purpose robots that operate seamlessly in any environment, with any object, and utilizing various skills to complete diverse tasks has been a long-standing goal in Artificial Intelligence. However, as a community, we have been constraining most robotic systems by designing them for specific tasks, training them on specific datasets, and deploying them within specific environments. These systems require extensively-labeled data and task-specific models. When deployed in real-world scenarios, such systems face several generalization issues and struggle to remain robust to distribution shifts. Motivated by the impressive open-set performance and content generation capabilities of web-scale, large-capacity pre-trained models (i.e., foundation models) in research fields such as Natural Language Processing (NLP) and Computer Vision (CV), we devote this survey to exploring (i) how these existing foundation models from NLP and CV can be applied to the field of general-purpose robotics, and also exploring (ii) what a robotics-specific foundation model would look like. We begin by providing a generalized formulation of how foundation models are used in robotics, and the fundamental barriers to making generalist robots universally applicable. Next, we establish a taxonomy to discuss current work exploring ways to leverage existing foundation models for robotics and develop ones catered to robotics. Finally, we discuss key challenges and promising future directions in using foundation models for enabling general-purpose robotic systems. We encourage readers to view our living GitHub repository 2 of resources, including papers reviewed in this survey, as well as related projects and repositories for developing foundation models for robotics.

cs.RO↗

arXiv2024

Robust Incremental Structure-from-Motion with Hybrid Features

Structure-from-Motion (SfM) has become a ubiquitous tool for camera calibration and scene reconstruction with many downstream applications in computer vision and beyond. While the state-of-the-art SfM pipelines have reached a high level of maturity in well-textured and well-configured scenes over the last decades, they still fall short of robustly solving the SfM problem in challenging scenarios. In particular, weakly textured scenes and poorly constrained configurations oftentimes cause catastrophic failures or large errors for the primarily keypoint-based pipelines. In these scenarios, line segments are often abundant and can offer complementary geometric constraints. Their large spatial extent and typically structured configurations lead to stronger geometric constraints as compared to traditional keypoint-based methods. In this work, we introduce an incremental SfM system that, in addition to points, leverages lines and their structured geometric relations. Our technical contributions span the entire pipeline (mapping, triangulation, registration) and we integrate these into a comprehensive end-to-end SfM system that we share as an open-source software with the community. We also present the first analytical method to propagate uncertainties for 3D optimized lines via sensitivity analysis. Experiments show that our system is consistently more robust and accurate compared to the widely used point-based state of the art in SfM -- achieving richer maps and more precise camera registrations, especially under challenging conditions. In addition, our uncertainty-aware localization module alone is able to consistently improve over the state of the art under both point-alone and hybrid setups.

cs.CV↗

arXiv2024

Teaching Robots Where To Go And How To Act With Human Sketches via Spatial Diagrammatic Instructions

This paper introduces Spatial Diagrammatic Instructions (SDIs), an approach for human operators to specify objectives and constraints that are related to spatial regions in the working environment. Human operators are enabled to sketch out regions directly on camera images that correspond to the objectives and constraints. These sketches are projected to 3D spatial coordinates, and continuous Spatial Instruction Maps (SIMs) are learned upon them. These maps can then be integrated into optimization problems for tasks of robots. In particular, we demonstrate how Spatial Diagrammatic Instructions can be applied to solve the Base Placement Problem of mobile manipulators, which concerns the best place to put the manipulator to facilitate a certain task. Human operators can specify, via sketch, spatial regions of interest for a manipulation task and permissible regions for the mobile manipulator to be at. Then, an optimization problem that maximizes the manipulator's reachability, or coverage, over the designated regions of interest while remaining in the permissible regions is solved. We provide extensive empirical evaluations, and show that our formulation of Spatial Instruction Maps provides accurate representations of user-specified diagrammatic instructions. Furthermore, we demonstrate that our diagrammatic approach to the Mobile Base Placement Problem enables higher quality solutions and faster runtime.

cs.RO↗

arXiv2024

Research on Predicting Public Opinion Event Heat Levels Based on Large Language Models

In recent years, with the rapid development of large language models, serval models such as GPT-4o have demonstrated extraordinary capabilities, surpassing human performance in various language tasks. As a result, many researchers have begun exploring their potential applications in the field of public opinion analysis. This study proposes a novel large-language-models-based method for public opinion event heat level prediction. First, we preprocessed and classified 62,836 Chinese hot event data collected between July 2022 and December 2023. Then, based on each event's online dissemination heat index, we used the MiniBatchKMeans algorithm to automatically cluster the events and categorize them into four heat levels (ranging from low heat to very high heat). Next, we randomly selected 250 events from each heat level, totalling 1,000 events, to build the evaluation dataset. During the evaluation process, we employed various large language models to assess their accuracy in predicting event heat levels in two scenarios: without reference cases and with similar case references. The results showed that GPT-4o and DeepseekV2 performed the best in the latter case, achieving prediction accuracies of 41.4% and 41.5%, respectively. Although the overall prediction accuracy remains relatively low, it is worth noting that for low-heat (Level 1) events, the prediction accuracies of these two models reached 73.6% and 70.4%, respectively. Additionally, the prediction accuracy showed a downward trend from Level 1 to Level 4, which correlates with the uneven distribution of data across the heat levels in the actual dataset. This suggests that with the more robust dataset, public opinion event heat level prediction based on large language models will have significant research potential for the future.

cs.CL↗

arXiv2024

CLSP: High-Fidelity Contrastive Language-State Pre-training for Agent State Representation

With the rapid development of artificial intelligence, multimodal learning has become an important research area. For intelligent agents, the state is a crucial modality to convey precise information alongside common modalities like images, videos, and language. This becomes especially clear with the broad adoption of reinforcement learning and multimodal large language models. Nevertheless, the representation of state modality still lags in development. To this end, we propose a High-Fidelity Contrastive Language-State Pre-training (CLSP) method, which can accurately encode state information into general representations for both reinforcement learning and multimodal large language models. Specifically, we first design a pre-training task based on the classification to train an encoder with coarse-grained information. Next, we construct data pairs of states and language descriptions, utilizing the pre-trained encoder to initialize the CLSP encoder. Then, we deploy contrastive learning to train the CLSP encoder to effectively represent precise state information. Additionally, we enhance the representation of numerical information using the Random Fourier Features (RFF) method for high-fidelity mapping. Extensive experiments demonstrate the superior precision and generalization capabilities of our representation, achieving outstanding results in text-state retrieval, reinforcement learning navigation tasks, and multimodal large language model understanding.

cs.AI↗