arXiv subjects

Zheng Yu

Publications and source records attributed to Zheng Yu.

At least 37 records · Page 2Linked to original sources

arXiv2024

GPTFUZZER: Red Teaming Large Language Models with Auto-Generated Jailbreak Prompts

Large language models (LLMs) have recently experienced tremendous popularity and are widely used from casual conversations to AI-driven programming. However, despite their considerable success, LLMs are not entirely reliable and can give detailed guidance on how to conduct harmful or illegal activities. While safety measures can reduce the risk of such outputs, adversarial jailbreak attacks can still exploit LLMs to produce harmful content. These jailbreak templates are typically manually crafted, making large-scale testing challenging. In this paper, we introduce GPTFuzz, a novel black-box jailbreak fuzzing framework inspired by the AFL fuzzing framework. Instead of manual engineering, GPTFuzz automates the generation of jailbreak templates for red-teaming LLMs. At its core, GPTFuzz starts with human-written templates as initial seeds, then mutates them to produce new templates. We detail three key components of GPTFuzz: a seed selection strategy for balancing efficiency and variability, mutate operators for creating semantically equivalent or similar sentences, and a judgment model to assess the success of a jailbreak attack. We evaluate GPTFuzz against various commercial and open-source LLMs, including ChatGPT, LLaMa-2, and Vicuna, under diverse attack scenarios. Our results indicate that GPTFuzz consistently produces jailbreak templates with a high success rate, surpassing human-crafted templates. Remarkably, GPTFuzz achieves over 90% attack success rates against ChatGPT and Llama-2 models, even with suboptimal initial seed templates. We anticipate that GPTFuzz will be instrumental for researchers and practitioners in examining LLM robustness and will encourage further exploration into enhancing LLM safety.

cs.AI↗

arXiv2024

CAMP: Compiler and Allocator-based Heap Memory Protection

The heap is a critical and widely used component of many applications. Due to its dynamic nature, combined with the complexity of heap management algorithms, it is also a frequent target for security exploits. To enhance the heap's security, various heap protection techniques have been introduced, but they either introduce significant runtime overhead or have limited protection. We present CAMP, a new sanitizer for detecting and capturing heap memory corruption. CAMP leverages a compiler and a customized memory allocator. The compiler adds boundary-checking and escape-tracking instructions to the target program, while the memory allocator tracks memory ranges, coordinates with the instrumentation, and neutralizes dangling pointers. With the novel error detection scheme, CAMP enables various compiler optimization strategies and thus eliminates redundant and unnecessary check instrumentation. This design minimizes runtime overhead without sacrificing security guarantees. Our evaluation and comparison of CAMP with existing tools, using both real-world applications and SPEC CPU benchmarks, show that it provides even better heap corruption detection capability with lower runtime overhead.

cs.CR↗

arXiv2024

Shortest paths govern fracture nucleation in thermoset networks

In this work, we introduce a predictive approach for determining fracture nucleation in thermosets based on shortest paths (SPs) of the network topology. This method enumerates SP sets in networks with periodic boundary conditions, with applications to both all-atom and coarse-grained simulations. We find that network fracture is most likely to nucleate on the first (shortest) SP and the strain at nucleation is linearly correlated with the topological path length. Subsequent fracture events are dictated by the instantaneous SP of partially broken networks. We further quantify the length scale dependence of SP distributions, introducing a means of bridging simulated and experimental fracture profiles.

cond-mat.soft↗

arXiv2024

How to Inverting the Leverage Score Distribution?

Leverage score is a fundamental problem in machine learning and theoretical computer science. It has extensive applications in regression analysis, randomized algorithms, and neural network inversion. Despite leverage scores are widely used as a tool, in this paper, we study a novel problem, namely the inverting leverage score problem. We analyze to invert the leverage score distributions back to recover model parameters. Specifically, given a leverage score $σ\in \mathbb{R}^n$, the matrix $A \in \mathbb{R}^{n \times d}$, and the vector $b \in \mathbb{R}^n$, we analyze the non-convex optimization problem of finding $x \in \mathbb{R}^d$ to minimize $\| \mathrm{diag}( σ) - I_n \circ (A(x) (A(x)^\top A(x) )^{-1} A(x)^\top ) \|_F$, where $A(x):= S(x)^{-1} A \in \mathbb{R}^{n \times d} $, $S(x) := \mathrm{diag}(s(x)) \in \mathbb{R}^{n \times n}$ and $s(x) : = Ax - b \in \mathbb{R}^n$. Our theoretical studies include computing the gradient and Hessian, demonstrating that the Hessian matrix is positive definite and Lipschitz, and constructing first-order and second-order algorithms to solve this regression problem. Our work combines iterative shrinking and the induction hypothesis to ensure global convergence rates for the Newton method, as well as the properties of Lipschitz and strong convexity to guarantee the performance of gradient descent. This important study on inverting statistical leverage opens up numerous new applications in interpretation, data recovery, and security.

cs.LG↗

arXiv2024

VL-Mamba: Exploring State Space Models for Multimodal Learning

Multimodal large language models (MLLMs) have attracted widespread interest and have rich applications. However, the inherent attention mechanism in its Transformer structure requires quadratic complexity and results in expensive computational overhead. Therefore, in this work, we propose VL-Mamba, a multimodal large language model based on state space models, which have been shown to have great potential for long-sequence modeling with fast inference and linear scaling in sequence length. Specifically, we first replace the transformer-based backbone language model such as LLama or Vicuna with the pre-trained Mamba language model. Then, we empirically explore how to effectively apply the 2D vision selective scan mechanism for multimodal learning and the combinations of different vision encoders and variants of pretrained Mamba language models. The extensive experiments on diverse multimodal benchmarks with competitive performance show the effectiveness of our proposed VL-Mamba and demonstrate the great potential of applying state space models for multimodal learning tasks.

cs.CV↗

arXiv2023

How close are the classical two-body potentials to ab initio calculations? Insights from linear machine learning based force matching

In this work, we propose a linear machine learning force matching approach that can directly extract pair atomic interactions from ab initio calculations in amorphous structures. The local feature representation is specifically chosen to make the linear weights a force field as a force/potential function of the atom pair distance. Consequently, this set of functions is the closest representation of the ab initio forces given the two-body approximation and finite scanning in the configurational space. We validate this approach in amorphous silica. Potentials in the new force field (consisting of tabulated Si-Si, Si-O, and O-O potentials) are significantly softer than existing potentials that are commonly used for silica, even though all of them produce the tetrahedral network structure and roughly similar glass properties. This suggests that those commonly used classical force fields do not offer fundamentally accurate representations of the atomic interaction in silica. The new force field furthermore produces a lower glass transition temperature ($T_g\sim$1800 K) and a positive liquid thermal expansion coefficient, suggesting the extraordinarily high $T_g$ and negative liquid thermal expansion of simulated silica could be artifacts of previously developed classical potentials. Overall, the proposed approach provides a fundamental yet intuitive way to evaluate two-body potentials against ab initio calculations, thereby offering an efficient way to guide the development of classical force fields.

cond-mat.mtrl-sci↗

arXiv2023

March in Chat: Interactive Prompting for Remote Embodied Referring Expression

Many Vision-and-Language Navigation (VLN) tasks have been proposed in recent years, from room-based to object-based and indoor to outdoor. The REVERIE (Remote Embodied Referring Expression) is interesting since it only provides high-level instructions to the agent, which are closer to human commands in practice. Nevertheless, this poses more challenges than other VLN tasks since it requires agents to infer a navigation plan only based on a short instruction. Large Language Models (LLMs) show great potential in robot action planning by providing proper prompts. Still, this strategy has not been explored under the REVERIE settings. There are several new challenges. For example, the LLM should be environment-aware so that the navigation plan can be adjusted based on the current visual observation. Moreover, the LLM planned actions should be adaptable to the much larger and more complex REVERIE environment. This paper proposes a March-in-Chat (MiC) model that can talk to the LLM on the fly and plan dynamically based on a newly proposed Room-and-Object Aware Scene Perceiver (ROASP). Our MiC model outperforms the previous state-of-the-art by large margins by SPL and RGSPL metrics on the REVERIE benchmark.

cs.CV↗

arXiv2023

VLN-PETL: Parameter-Efficient Transfer Learning for Vision-and-Language Navigation

The performance of the Vision-and-Language Navigation~(VLN) tasks has witnessed rapid progress recently thanks to the use of large pre-trained vision-and-language models. However, full fine-tuning the pre-trained model for every downstream VLN task is becoming costly due to the considerable model size. Recent research hotspot of Parameter-Efficient Transfer Learning (PETL) shows great potential in efficiently tuning large pre-trained models for the common CV and NLP tasks, which exploits the most of the representation knowledge implied in the pre-trained model while only tunes a minimal set of parameters. However, simply utilizing existing PETL methods for the more challenging VLN tasks may bring non-trivial degeneration to the performance. Therefore, we present the first study to explore PETL methods for VLN tasks and propose a VLN-specific PETL method named VLN-PETL. Specifically, we design two PETL modules: Historical Interaction Booster (HIB) and Cross-modal Interaction Booster (CIB). Then we combine these two modules with several existing PETL methods as the integrated VLN-PETL. Extensive experimental results on four mainstream VLN tasks (R2R, REVERIE, NDH, RxR) demonstrate the effectiveness of our proposed VLN-PETL, where VLN-PETL achieves comparable or even better performance to full fine-tuning and outperforms other PETL methods with promising margins.

cs.CV↗

arXiv2023

Sketching for First Order Method: Efficient Algorithm for Low-Bandwidth Channel and Vulnerability

Sketching is one of the most fundamental tools in large-scale machine learning. It enables runtime and memory saving via randomly compressing the original large problem into lower dimensions. In this paper, we propose a novel sketching scheme for the first order method in large-scale distributed learning setting, such that the communication costs between distributed agents are saved while the convergence of the algorithms is still guaranteed. Given gradient information in a high dimension $d$, the agent passes the compressed information processed by a sketching matrix $R\in \mathbb{R}^{s\times d}$ with $s\ll d$, and the receiver de-compressed via the de-sketching matrix $R^\top$ to ``recover'' the information in original dimension. Using such a framework, we develop algorithms for federated learning with lower communication costs. However, such random sketching does not protect the privacy of local data directly. We show that the gradient leakage problem still exists after applying the sketching technique by presenting a specific gradient attack method. As a remedy, we prove rigorously that the algorithm will be differentially private by adding additional random noises in gradient information, which results in a both communication-efficient and differentially private first order approach for federated learning tasks. Our sketching scheme can be further generalized to other learning settings and might be of independent interest itself.

cs.LG↗

arXiv2023

Deep Reinforcement Learning for Cost-Effective Medical Diagnosis

Dynamic diagnosis is desirable when medical tests are costly or time-consuming. In this work, we use reinforcement learning (RL) to find a dynamic policy that selects lab test panels sequentially based on previous observations, ensuring accurate testing at a low cost. Clinical diagnostic data are often highly imbalanced; therefore, we aim to maximize the $F_1$ score instead of the error rate. However, optimizing the non-concave $F_1$ score is not a classic RL problem, thus invalidates standard RL methods. To remedy this issue, we develop a reward shaping approach, leveraging properties of the $F_1$ score and duality of policy optimization, to provably find the set of all Pareto-optimal policies for budget-constrained $F_1$ score maximization. To handle the combinatorially complex state space, we propose a Semi-Model-based Deep Diagnosis Policy Optimization (SM-DDPO) framework that is compatible with end-to-end training and online learning. SM-DDPO is tested on diverse clinical tasks: ferritin abnormality detection, sepsis mortality prediction, and acute kidney injury diagnosis. Experiments with real-world data validate that SM-DDPO trains efficiently and identifies all Pareto-front solutions. Across all tasks, SM-DDPO is able to achieve state-of-the-art diagnosis accuracy (in some cases higher than conventional methods) with up to $85\%$ reduction in testing cost. The code is available at [https://github.com/Zheng321/Deep-Reinforcement-Learning-for-Cost-Effective-Medical-Diagnosis].

cs.LG↗

arXiv2022

HOP: History-and-Order Aware Pre-training for Vision-and-Language Navigation

Pre-training has been adopted in a few of recent works for Vision-and-Language Navigation (VLN). However, previous pre-training methods for VLN either lack the ability to predict future actions or ignore the trajectory contexts, which are essential for a greedy navigation process. In this work, to promote the learning of spatio-temporal visual-textual correspondence as well as the agent's capability of decision making, we propose a novel history-and-order aware pre-training paradigm (HOP) with VLN-specific objectives that exploit the past observations and support future action prediction. Specifically, in addition to the commonly used Masked Language Modeling (MLM) and Trajectory-Instruction Matching (TIM), we design two proxy tasks to model temporal order information: Trajectory Order Modeling (TOM) and Group Order Modeling (GOM). Moreover, our navigation action prediction is also enhanced by introducing the task of Action Prediction with History (APH), which takes into account the history visual perceptions. Extensive experimental results on four downstream VLN tasks (R2R, REVERIE, NDH, RxR) demonstrate the effectiveness of our proposed method compared against several state-of-the-art agents.

cs.CV↗

arXiv2022

Understanding the fragile-to-strong transition in silica from microscopic dynamics

In this work, we revisit the fragile-to-strong transition (FTS) in the simulated BKS silica from the perspective of microscopic dynamics in an effort to elucidate the dynamical behaviors of fragile and strong glass-forming liquids. Softness, which is a machine-learned feature from local atomic structures, is used to predict the microscopic activation energetics and long-term dynamics. The FTS is found to originate from a change in the temperature dependence of the microscopic activation energetics. Furthermore, results suggest there are two diffusion channels with different energy barriers in BKS silica. The fast dynamics at high temperatures is dominated by the channel with small energy barriers ($<\sim$1 eV), which is controlled by the short-range order. The rapid closing of this diffusion channel when lowering temperature leads to the fragile behavior. On the other hand, the slow dynamics at low temperatures is dominated by the channel with large energy barriers controlled by the medium-range order. This slow diffusion channel changes only subtly with temperature, leading to the strong behavior. The distributions of barriers in the two channels show different temperature dependences, causing a crossover at $\sim$3100 K. This transition temperature in microscopic dynamics is consistent with the inflection point in the configurational entropy, suggesting there is a fundamental correlation between microscopic dynamics and thermodynamics.

cond-mat.dis-nn↗

arXiv2021

Fast Graph Neural Tangent Kernel via Kronecker Sketching

Many deep learning tasks have to deal with graphs (e.g., protein structures, social networks, source code abstract syntax trees). Due to the importance of these tasks, people turned to Graph Neural Networks (GNNs) as the de facto method for learning on graphs. GNNs have become widely applied due to their convincing performance. Unfortunately, one major barrier to using GNNs is that GNNs require substantial time and resources to train. Recently, a new method for learning on graph data is Graph Neural Tangent Kernel (GNTK) [Du, Hou, Salakhutdinov, Poczos, Wang and Xu 19]. GNTK is an application of Neural Tangent Kernel (NTK) [Jacot, Gabriel and Hongler 18] (a kernel method) on graph data, and solving NTK regression is equivalent to using gradient descent to train an infinite-wide neural network. The key benefit of using GNTK is that, similar to any kernel method, GNTK's parameters can be solved directly in a single step. This can avoid time-consuming gradient descent. Meanwhile, sketching has become increasingly used in speeding up various optimization problems, including solving kernel regression. Given a kernel matrix of $n$ graphs, using sketching in solving kernel regression can reduce the running time to $o(n^3)$. But unfortunately such methods usually require extensive knowledge about the kernel matrix beforehand, while in the case of GNTK we find that the construction of the kernel matrix is already $O(n^2N^4)$, assuming each graph has $N$ nodes. The kernel matrix construction time can be a major performance bottleneck when the size of graphs $N$ increases. A natural question to ask is thus whether we can speed up the kernel matrix construction to improve GNTK regression's end-to-end running time. This paper provides the first algorithm to construct the kernel matrix in $o(n^2N^3)$ running time.

cs.LG↗

arXiv2021

Fast Sketching of Polynomial Kernels of Polynomial Degree

Kernel methods are fundamental in machine learning, and faster algorithms for kernel approximation provide direct speedups for many core tasks in machine learning. The polynomial kernel is especially important as other kernels can often be approximated by the polynomial kernel via a Taylor series expansion. Recent techniques in oblivious sketching reduce the dependence in the running time on the degree $q$ of the polynomial kernel from exponential to polynomial, which is useful for the Gaussian kernel, for which $q$ can be chosen to be polylogarithmic. However, for more slowly growing kernels, such as the neural tangent and arc-cosine kernels, $q$ needs to be polynomial, and previous work incurs a polynomial factor slowdown in the running time. We give a new oblivious sketch which greatly improves upon this running time, by removing the dependence on $q$ in the leading order term. Combined with a novel sampling scheme, we give the fastest algorithms for approximating a large family of slow-growing kernels.

cs.DS↗

arXiv2021

On the Convergence and Sample Efficiency of Variance-Reduced Policy Gradient Method

Policy gradient (PG) gives rise to a rich class of reinforcement learning (RL) methods. Recently, there has been an emerging trend to accelerate the existing PG methods such as REINFORCE by the \emph{variance reduction} techniques. However, all existing variance-reduced PG methods heavily rely on an uncheckable importance weight assumption made for every single iteration of the algorithms. In this paper, a simple gradient truncation mechanism is proposed to address this issue. Moreover, we design a Truncated Stochastic Incremental Variance-Reduced Policy Gradient (TSIVR-PG) method, which is able to maximize not only a cumulative sum of rewards but also a general utility function over a policy's long-term visiting distribution. We show an $\tilde{\mathcal{O}}(ε^{-3})$ sample complexity for TSIVR-PG to find an $ε$-stationary policy. By assuming the overparameterizaiton of policy and exploiting the hidden convexity of the problem, we further show that TSIVR-PG converges to global $ε$-optimal policy with $\tilde{\mathcal{O}}(ε^{-2})$ samples.

cs.LG↗

arXiv2021

Mapping the Galactic Disk with the LAMOST and Gaia Red Clump Sample VII: the Stellar Disk Structure Revealed by the Mono-abundance Populations

Using a sample of 96,201 primary red clump (RC) stars selected from the LAMOST and Gaia surveys, we investigate the stellar structure of the Galactic disk. The sample stars show two separated sequences of high-[α/Fe] and low-[α/Fe] in the [α/Fe]-[Fe/H] plane. We divide the sample stars into five mono-abundance populations (MAPs) with different ranges of [α/Fe] and [Fe/H], named as the high-[α/Fe], high-[α/Fe] & high-[Fe/H], low-[Fe/H], solar, high-[Fe/H] MAPs respectively. We present the stellar number density distributions in the R R Z plane, and the scale heights and scale lengths of the individual MAPs by fitting their vertical and radial density profiles. The vertical profiles, the variation trend of scale height with the Galactocentric radius, indicate that there is a clear disk flare in the outer disk both for the low-[α/Fe] and the high-[α/Fe] MAPs. While the radial surface density profiles show a peak radius of 7 kpc and 8 kpc for the high-[α/Fe] and low-[α/Fe] MAPs, respectively. We also investigate the correlation between the mean rotation velocity and metallicity of the individual MAPs, and find that the mean rotation velocities are well separated and show different trends between the high-[α/Fe] and the low-[α/Fe] MAPs. At last, we discuss the character of the high-[α/Fe] & high-[Fe/H] MAP and find that it is more similar to the high-[α/Fe] MAP either in the radial and vertical density profiles or in the rotation velocity.

astro-ph.GA↗

arXiv2020

Structural Signatures for Thermodynamic Stability in Vitreous Silica: Insight from Machine Learning and Molecular Dynamics Simulations

The structure-thermodynamic stability relationship in vitreous silica is investigated using machine learning and a library of 24,157 inherent structures generated from melt-quenching and replica exchange molecular dynamics simulations. We find the thermodynamic stability, i.e., enthalpy of the inherent structure ($e_{\mathrm{IS}}$), can be accurately predicted by both linear and nonlinear machine learning models from numeric structural descriptors commonly used to characterize disordered structures. We find short-range features become less indicative of thermodynamic stability below the fragile-to-strong transition. On the other hand, medium-range features, especially those between 2.8-~6 $\unicode{x212B}$;, show consistent correlations with $e_{\mathrm{IS}}$ across the liquid and glass regions, and are found to be the most critical to stability prediction among features from different length scales. Based on the machine learning models, a set of five structural features that are the most predictive of the silica glass stability is identified.

cond-mat.dis-nn↗

arXiv2020

Generalized Leverage Score Sampling for Neural Networks

Leverage score sampling is a powerful technique that originates from theoretical computer science, which can be used to speed up a large number of fundamental questions, e.g. linear regression, linear programming, semi-definite programming, cutting plane method, graph sparsification, maximum matching and max-flow. Recently, it has been shown that leverage score sampling helps to accelerate kernel methods [Avron, Kapralov, Musco, Musco, Velingker and Zandieh 17]. In this work, we generalize the results in [Avron, Kapralov, Musco, Musco, Velingker and Zandieh 17] to a broader class of kernels. We further bring the leverage score sampling into the field of deep learning theory. $\bullet$ We show the connection between the initialization for neural network training and approximating the neural tangent kernel with random features. $\bullet$ We prove the equivalence between regularized neural network and neural tangent kernel ridge regression under the initialization of both classical random Gaussian and leverage score sampling.

cs.LG↗