arXiv subjects

Yichen Xu

Publications and source records attributed to Yichen Xu.

At least 55 records · Page 3Linked to original sources

arXiv2024

Hierarchical Diffusion Policy: manipulation trajectory generation via contact guidance

Decision-making in robotics using denoising diffusion processes has increasingly become a hot research topic, but end-to-end policies perform poorly in tasks with rich contact and have limited controllability. This paper proposes Hierarchical Diffusion Policy (HDP), a new imitation learning method of using objective contacts to guide the generation of robot trajectories. The policy is divided into two layers: the high-level policy predicts the contact for the robot's next object manipulation based on 3D information, while the low-level policy predicts the action sequence toward the high-level contact based on the latent variables of observation and contact. We represent both level policies as conditional denoising diffusion processes, and combine behavioral cloning and Q-learning to optimize the low level policy for accurately guiding actions towards contact. We benchmark Hierarchical Diffusion Policy across 6 different tasks and find that it significantly outperforms the existing state of-the-art imitation learning method Diffusion Policy with an average improvement of 20.8%. We find that contact guidance yields significant improvements, including superior performance, greater interpretability, and stronger controllability, especially on contact-rich tasks. To further unlock the potential of HDP, this paper proposes a set of key technical contributions including snapshot gradient optimization, 3D conditioning, and prompt guidance, which improve the policy's optimization efficiency, spatial awareness, and controllability respectively. Finally, real world experiments verify that HDP can handle both rigid and deformable objects.

cs.RO↗

arXiv2024

Attention to Quantum Complexity

The imminent era of error-corrected quantum computing urgently demands robust methods to characterize complex quantum states, even from limited and noisy measurements. We introduce the Quantum Attention Network (QuAN), a versatile classical AI framework leveraging the power of attention mechanisms specifically tailored to address the unique challenges of learning quantum complexity. Inspired by large language models, QuAN treats measurement snapshots as tokens while respecting their permutation invariance. Combined with a novel parameter-efficient mini-set self-attention block (MSSAB), such data structure enables QuAN to access high-order moments of the bit-string distribution and preferentially attend to less noisy snapshots. We rigorously test QuAN across three distinct quantum simulation settings: driven hard-core Bose-Hubbard model, random quantum circuits, and the toric code under coherent and incoherent noise. QuAN directly learns the growth in entanglement and state complexity from experimentally obtained computational basis measurements. In particular, it learns the growth in complexity of random circuit data upon increasing depth from noisy experimental data. Taken to a regime inaccessible by existing theory, QuAN unveils the complete phase diagram for noisy toric code data as a function of both noise types. This breakthrough highlights the transformative potential of using purposefully designed AI-driven solutions to assist quantum hardware.

quant-ph↗

arXiv2024

Fluctuation-Dissipation Theorem and Information Geometry in Open Quantum Systems

We propose a fluctuation-dissipation theorem in open quantum systems from an information-theoretic perspective. We define the fidelity susceptibility that measures the sensitivity of the systems under perturbation and relate it to the fidelity correlator that characterizes the correlation behaviors for mixed quantum states. In particular, we determine the scaling behavior of the fidelity susceptibility in the strong-to-weak spontaneous symmetry breaking (SW-SSB) phase, strongly symmetric short-range correlated phase, and the quantum critical point between them. We then provide a geometric perspective of our construction using distance measures of density matrices. We find that the metric of the quantum information geometry generated by perturbative distance between density matrices before and after perturbation is generally non-analytic. Finally, we design a polynomial proxy that can in principle be used as an experimental probe for detecting the SW-SSB and phase transition through quantum metrology. In particular, we show that each term of the polynomial proxy is related to the Rényi versions of the fidelity correlators.

quant-ph↗

arXiv2024

TinyChart: Efficient Chart Understanding with Visual Token Merging and Program-of-Thoughts Learning

Charts are important for presenting and explaining complex data relationships. Recently, multimodal large language models (MLLMs) have shown remarkable capabilities in various chart understanding tasks. However, the sheer size of these models in terms of parameters and computational requirements limits their use in resource-constrained environments. In this paper, we present TinyChart, an efficient MLLM for chart understanding with only 3B parameters. TinyChart overcomes two key challenges in efficient chart understanding: (1) reduce the burden of learning numerical computations through a Program-of-Thoughts (PoT) learning strategy, which trains the model to generate Python programs for numerical calculations, and (2) reduce lengthy vision feature sequences produced by the vision transformer for high-resolution images through a Vision Token Merging module, which gradually merges most similar vision tokens. Extensive experiments demonstrate that our 3B TinyChart achieves SOTA performance on a variety of chart understanding benchmarks including ChartQA, Chart-to-Text, Chart-to-Table, OpenCQA, and ChartX. It outperforms several chart understanding MLLM with up to 13B parameters such as ChartLlama and ChartAst, and close-sourced general-purpose MLLM GPT-4V on ChartQA. It also demonstrates its superior efficiency with higher throughput during inference due to a smaller model scale and more efficient vision encoding. Our code and model are available at https://github.com/X-PLUG/mPLUG-DocOwl/tree/main/TinyChart.

cs.CV↗

arXiv2024

VersaT2I: Improving Text-to-Image Models with Versatile Reward

Recent text-to-image (T2I) models have benefited from large-scale and high-quality data, demonstrating impressive performance. However, these T2I models still struggle to produce images that are aesthetically pleasing, geometrically accurate, faithful to text, and of good low-level quality. We present VersaT2I, a versatile training framework that can boost the performance with multiple rewards of any T2I model. We decompose the quality of the image into several aspects such as aesthetics, text-image alignment, geometry, low-level quality, etc. Then, for every quality aspect, we select high-quality images in this aspect generated by the model as the training set to finetune the T2I model using the Low-Rank Adaptation (LoRA). Furthermore, we introduce a gating function to combine multiple quality aspects, which can avoid conflicts between different quality aspects. Our method is easy to extend and does not require any manual annotation, reinforcement learning, or model architecture changes. Extensive experiments demonstrate that VersaT2I outperforms the baseline methods across various quality criteria.

cs.CV↗

arXiv2024

URHand: Universal Relightable Hands

Existing photorealistic relightable hand models require extensive identity-specific observations in different views, poses, and illuminations, and face challenges in generalizing to natural illuminations and novel identities. To bridge this gap, we present URHand, the first universal relightable hand model that generalizes across viewpoints, poses, illuminations, and identities. Our model allows few-shot personalization using images captured with a mobile phone, and is ready to be photorealistically rendered under novel illuminations. To simplify the personalization process while retaining photorealism, we build a powerful universal relightable prior based on neural relighting from multi-view images of hands captured in a light stage with hundreds of identities. The key challenge is scaling the cross-identity training while maintaining personalized fidelity and sharp details without compromising generalization under natural illuminations. To this end, we propose a spatially varying linear lighting model as the neural renderer that takes physics-inspired shading as input feature. By removing non-linear activations and bias, our specifically designed lighting model explicitly keeps the linearity of light transport. This enables single-stage training from light-stage data while generalizing to real-time rendering under arbitrary continuous illuminations across diverse identities. In addition, we introduce the joint learning of a physically based model and our neural relighting model, which further improves fidelity and generalization. Extensive experiments show that our approach achieves superior performance over existing methods in terms of both quality and generalizability. We also demonstrate quick personalization of URHand from a short phone scan of an unseen identity.

cs.CV↗

arXiv2023

Devil in the Number: Towards Robust Multi-modality Data Filter

In order to appropriately filter multi-modality data sets on a web-scale, it becomes crucial to employ suitable filtering methods to boost performance and reduce training costs. For instance, LAION papers employs the CLIP score filter to select data with CLIP scores surpassing a certain threshold. On the other hand, T-MARS achieves high-quality data filtering by detecting and masking text within images and then filtering by CLIP score. Through analyzing the dataset, we observe a significant proportion of redundant information, such as numbers, present in the textual content. Our experiments on a subset of the data unveil the profound impact of these redundant elements on the CLIP scores. A logical approach would involve reevaluating the CLIP scores after eliminating these influences. Experimentally, our text-based CLIP filter outperforms the top-ranked method on the ``small scale" of DataComp (a data filtering benchmark) on ImageNet distribution shifts, achieving a 3.6% performance improvement. The results also demonstrate that our proposed text-masked filter outperforms the original CLIP score filter when selecting the top 40% of the data. The impact of numbers on CLIP and their handling provide valuable insights for improving the effectiveness of CLIP training, including language rewrite techniques.

cs.LG↗

arXiv2023

Degrees of Separation: A Flexible Type System for Data Race Prevention

Data races are a notorious problem in parallel programming. There has been great research interest in type systems that statically prevent data races. Despite the progress in the safety and usability of these systems, lots of existing approaches enforce strict anti-aliasing principles to prevent data races. The adoption of them is often intrusive, in the sense that it invalidates common programming patterns and requires paradigm shifts. We propose Capture Separation Calculus (System CSC), a calculus based on Capture Calculus (System CC<:box), that achieves static data race freedom while being non-intrusive. It allows aliasing in general to permit common programming patterns, but tracks aliasing and controls them when that is necessary to prevent data races. We study the formal properties of System CSC by establishing its type safety and data race freedom. Notably, we establish the data race freedom property by proving the confluence of its reduction semantics. To validate the usability of the calculus, we implement it as an extension to the Scala 3 compiler, and use it to type-check the examples in the paper.

cs.PL↗

arXiv2023

A Mechanized Theory of the Box Calculus

The capture calculus is an extension of System F<: that tracks free variables of terms in their type, allowing one to represent capabilities while limiting their scope. While previous calculi had mechanized soundness proofs -- notably System CF<: -- the latest version, namely the box calculus (System CC<:box), only had a paper proof. We present here our work on mechanizing the theory of the box calculus in Coq, and the challenges encountered along the way. While doing so, we motivate the current design of capture calculus, in particular the concept of boxes, from both user and metatheoretical standpoints. Our mechanization is complete and available on GitHub.

cs.LO↗

arXiv2023

Formalizing Box Inference for Capture Calculus

Capture calculus has recently been proposed as a solution to effect checking, achieved by tracking the captured references of terms in the types. Boxes, along with the box and unbox operations, are a crucial construct in capture calculus, which maintains the hygiene of types and improves the expressiveness of polymorphism over capturing types. Despite their usefulness in the formalism, boxes would soon become a heavy notational burden for users when the program grows. It thus necessitates the inference of boxes when integrating capture checking into a mainstream programming language. In this paper, we develop a formalisation of box inference for capture calculus. We begin by introducing a semi-algorithmic variant of the capture calculus, from which we derive an inference system where typed transformations are applied to complete missing box operations in programs. Then, we propose a type-level system that performs provably equivalent inference on the type level, without actually transforming the program. In the metatheory, we establish the relationships between these systems and capture calculus, thereby proving both the soundness and the completeness of box inference.

cs.PL↗

arXiv2023

A Secure and Efficient Multi-Object Grasping Detection Approach for Robotic Arms

Robotic arms are widely used in automatic industries. However, with wide applications of deep learning in robotic arms, there are new challenges such as the allocation of grasping computing power and the growing demand for security. In this work, we propose a robotic arm grasping approach based on deep learning and edge-cloud collaboration. This approach realizes the arbitrary grasp planning of the robot arm and considers the grasp efficiency and information security. In addition, the encoder and decoder trained by GAN enable the images to be encrypted while compressing, which ensures the security of privacy. The model achieves 92% accuracy on the OCID dataset, the image compression ratio reaches 0.03%, and the structural difference value is higher than 0.91.

cs.RO↗

arXiv2023

A case for DOT: Theoretical Foundations for Objects With Pattern Matching and GADT-style Reasoning

Many programming languages in the OO tradition now support pattern matching in some form. Historical examples include Scala and Ceylon, with the more recent additions of Java, Kotlin, TypeScript, and Flow. But pattern matching on generic class hierarchies currently results in puzzling type errors in most of these languages. Yet this combination of features occurs naturally in many scenarios, such as when manipulating typed ASTs. To support it properly, compilers need to implement a form of subtyping reconstruction: the ability to reconstruct subtyping information uncovered at runtime during pattern matching. We introduce cDOT, a new calculus in the family of Dependent Object Types (DOT) intended to serve as a formal foundation for subtyping reconstruction. Being descended from pDOT, itself a formal foundation for Scala, cDOT can be used to encode advanced object-oriented features such as generic inheritance, type constructor variance, F-bounded polymorphism, and first-class recursive modules. We demonstrate that subtyping reconstruction subsumes GADTs by encoding $λ_{2,Gμ}$, a classical constraint-based GADT calculus, into cDOT.

cs.PL↗

arXiv2022

A Survey on Pretrained Language Models for Neural Code Intelligence

As the complexity of modern software continues to escalate, software engineering has become an increasingly daunting and error-prone endeavor. In recent years, the field of Neural Code Intelligence (NCI) has emerged as a promising solution, leveraging the power of deep learning techniques to tackle analytical tasks on source code with the goal of improving programming efficiency and minimizing human errors within the software industry. Pretrained language models have become a dominant force in NCI research, consistently delivering state-of-the-art results across a wide range of tasks, including code summarization, generation, and translation. In this paper, we present a comprehensive survey of the NCI domain, including a thorough review of pretraining techniques, tasks, datasets, and model architectures. We hope this paper will serve as a bridge between the natural language and programming language communities, offering insights for future research in this rapidly evolving field.

cs.SE↗

arXiv2022

A Survey on Deep Graph Generation: Methods and Applications

Graphs are ubiquitous in encoding relational information of real-world objects in many domains. Graph generation, whose purpose is to generate new graphs from a distribution similar to the observed graphs, has received increasing attention thanks to the recent advances of deep learning models. In this paper, we conduct a comprehensive review on the existing literature of deep graph generation from a variety of emerging methods to its wide application areas. Specifically, we first formulate the problem of deep graph generation and discuss its difference with several related graph learning tasks. Secondly, we divide the state-of-the-art methods into three categories based on model architectures and summarize their generation strategies. Thirdly, we introduce three key application areas of deep graph generation. Lastly, we highlight challenges and opportunities in the future study of deep graph generation. We hope that our survey will be useful for researchers and practitioners who are interested in this exciting and rapidly-developing field.

cs.LG↗

arXiv2022

Deconfined Quantum Critical Point with Non-locality

The deconfined quantum critical point (DQCP) between the Néel and valence bond solid (VBS) order was originally proposed in quantum spin systems with a local Hamiltonian. In the last few years analogues of DQCPs with nonlocal interactions have been explored, which can lead to rich possibilities. The nonlocal interactions can either arise from an instantaneous long range interaction in the Hamiltonian, or from gapless modes that reside in one higher spatial dimension. Here we consider another mechanism of generating nonlocal interactions by coupling the DQCP to the "hot spots" of a Fermi surface. We demonstrate that at least within a substantial energy window, the physics of the DQCP is controlled by a new fixed point with dynamical exponent $z > 1$

cond-mat.str-el↗

arXiv2022

Metal-Insulator Transition with Charge Fractionalization

It has been proposed that an extended version of the Hubbard model which potentially hosts rich correlated physics may be well simulated by the transition metal dichalcogenide (TMD) moiré heterostructures. Motivated by recent reports of continuous metal-insulator transition (MIT) at half filling, as well as correlated insulators at various fractional fillings in TMD moiré heterostructures, we propose a theory for the potentially continuous MIT with fractionalized electric charges. The charge fractionalization at the MIT will lead to various experimental observable effects, such as a large critical resistivity as well as large universal resistivity jump at the continuous MIT. These predictions are different from previously proposed theory for interaction-driven continuous MIT. Physics in phases near the MIT will also be discussed.

cond-mat.str-el↗

arXiv2022

A construction of exotic metallic states

We discuss examples of two dimensional metallic states with charge fractionalization, and we will demonstrate that the mechanism of charge fractionalization leads to exotic metallic behaviors at low and intermediate temperature. The simplest example of such state is constructed by fermionic partons at finite density coupled to a $Z_N$ gauge field, whose properties can be studied through rudimentary methods. This simple state has the following exotic features: (1) at low temperature this state is a "bad metal" whose resistivity can exceed the Mott-Ioffe-Regel limit; (2) while increasing temperature $T$ the resistivity $ρ(T)$ is a nonmonotonic function, and it crosses over from a bad metal at low $T$ to a good metal at relatively high $T$; (3) the optical conductivity $σ(ω)$ has a small Drude weight at low $T$, and a larger Drude weight at intermediate $T$; (4) at low temperature the metallic state has a large Lorenz number, which strongly violates the Wiedemann-Franz law. A more complex example with fermionic partons at finite density coupled to a SU(N) gauge field will also be constructed.

cond-mat.str-el↗

arXiv2021

An Empirical Study of Graph Contrastive Learning

Graph Contrastive Learning (GCL) establishes a new paradigm for learning graph representations without human annotations. Although remarkable progress has been witnessed recently, the success behind GCL is still left somewhat mysterious. In this work, we first identify several critical design considerations within a general GCL paradigm, including augmentation functions, contrasting modes, contrastive objectives, and negative mining techniques. Then, to understand the interplay of different GCL components, we conduct extensive, controlled experiments over a set of benchmark tasks on datasets across various domains. Our empirical studies suggest a set of general receipts for effective GCL, e.g., simple topology augmentations that produce sparse graph views bring promising performance improvements; contrasting modes should be aligned with the granularities of end tasks. In addition, to foster future research and ease the implementation of GCL algorithms, we develop an easy-to-use library PyGCL, featuring modularized CL components, standardized evaluation, and experiment management. We envision this work to provide useful empirical evidence of effective GCL algorithms and offer several insights for future research.

cs.LG↗