Search

With the growing interest in Quantum Machine Learning (QML) and the increasing availability of quantum computers through cloud providers, addressing the potential security risks associated with QML has become an urgent priority. One key concern in the QML domain is the threat of data poisoning attacks in the current quantum cloud setting. Adversarial access to training data could severely compromise the integrity and availability of QML models. Classical data poisoning techniques require significant knowledge and training to generate poisoned data, and lack noise resilience, making them ineffective for QML models in the Noisy Intermediate Scale Quantum (NISQ) era. In this work, we first propose a simple yet effective technique to measure intra-class encoder state similarity (ESS) by analyzing the outputs of encoding circuits. Leveraging this approach, we introduce a quantum indiscriminate data poisoning attack, QUID. Through extensive experiments conducted in both noiseless and noisy environments (e.g., IBM\_Brisbane's noise), across various architectures and datasets, QUID achieves up to $92\%$ accuracy degradation in model performance compared to baseline models and up to $75\%$ accuracy degradation compared to random label-flipping. We also tested QUID against state-of-the-art classical defenses, with accuracy degradation still exceeding $50\%$, demonstrating its effectiveness. This work represents the first attempt to reevaluate data poisoning attacks in the context of QML.

Quantum error correction (QEC) is crucial for ensuring the reliability of quantum computers. However, implementing QEC often requires a significant number of qubits, leading to substantial overhead. One of the major challenges in quantum computing is reducing this overhead, especially since QEC codes depend heavily on ancilla qubits for stabilizer measurements. In this work, we propose reducing the number of ancilla qubits by reusing the same ancilla qubits for both X- and Z-type stabilizers. This is achieved by alternating between X and Z stabilizer measurements during each half-round, cutting the number of required ancilla qubits in half. This technique can be applied broadly across various QEC codes, we focus on rotated surface codes only and achieve nearly \(25\%\) reduction in total qubit overhead. We also present a few use cases where the proposed idea enables the usage of higher-distance surface codes at a relatively lesser qubit count. Our analysis shows that the modified approach enables users to achieve similar or better error correction with fewer qubits, especially for higher distances (\(d \geq 13\)). Additionally, we identify conditions where the modified code allows for extended distances (\(d + k\)) while using the same or fewer resources as the original, offering a scalable and practical solution for quantum error correction. These findings emphasize the modified surface code's potential to optimize qubit usage in resource-constrained quantum systems., Comment: 7 pages, 4 figures, 3 tables

Thought experiments are considered valuable tools in science, enabling the exploration of hypotheses and the examination of complex ideas in a conceptual, non-empirical framework. These thought experiments can be useful in design fiction for speculating future possibilities, examining existing and alternate scenarios in new ways or challenging current paradigms. In visualization, speculating future possibilities or exploring new ways of interpreting existing scenarios can provoke critical reflection and envision novel approaches. In this paper we present such thought experiments for visualization. We conceptualize and define a thought experiment to consist of a situation, a story, and a scenario. Situations are derived from different tools of thought experiments and visualization practice; a story is an AI-generated fiction based on the situation and the scenario is the grounding of the situation and story in visualization research. We present ten such thought experiments and demonstrate their utility in visualization by deriving critiques from them.

This paper addresses the lateral control of Autonomous and Connected Vehicles (ACVs) in a platoon executing an Emergency Lane Change (ELC) maneuver. These maneuvers are typically triggered by emergency signals from the front or rear of the platoon in response to the need to avoid obstacles or allow other vehicles to pass. The study assumes that ACVs maintain reliable connectivity, enabling each following vehicle to access GPS position traces of both the lead and immediately preceding vehicles in the platoon. We demonstrate that lateral string stability in the ACV platoon can be achieved using communicated information solely from the lead and preceding vehicles. Additionally, we present a lateral control framework for ACVs, which helps track a discretized preview of the trajectory constructed from the communicated data. This framework involves constructing two distinct trajectories based on the preview data from the lead and preceding vehicles, calculating the associated errors and lateral control actions for each, and then integrating these to generate a steering command. Numerical results validate the effectiveness of the proposed lateral control scheme., Comment: 18th IEEE International Conference on Vehicular Electronics and Safety 2024 (ICVES)

Federated Learning (FL) is a form of distributed learning that allows multiple institutions or clients to collaboratively learn a global model to solve a task. This allows the model to utilize the information from every institute while preserving data privacy. However, recent studies show that the promise of protecting the privacy of data is not upheld by existing methods and that it is possible to recreate the training data from the different institutions. This is done by utilizing gradients transferred between the clients and the global server during training or by knowing the model architecture at the client end. In this paper, we propose a federated learning framework for semantic segmentation without knowing the model architecture nor transferring gradients between the client and the server, thus enabling better privacy preservation. We propose BlackFed - a black-box adaptation of neural networks that utilizes zero order optimization (ZOO) to update the client model weights and first order optimization (FOO) to update the server weights. We evaluate our approach on several computer vision and medical imaging datasets to demonstrate its effectiveness. To the best of our knowledge, this work is one of the first works in employing federated learning for segmentation, devoid of gradients or model information exchange. Code: https://github.com/JayParanjape/blackfed/tree/master, Comment: Accepted at NEURIPS 2024

In this paper, a heuristic for a heterogeneous min-max multi-vehicle multi-depot traveling salesman problem is proposed, wherein heterogeneous vehicles start from given depot locations and need to cover a given set of targets. In the considered problem, vehicles can be structurally heterogeneous due to different vehicle speeds and/or functionally heterogeneous due to different vehicle-target assignments originating from different sensing capabilities of vehicles. The proposed heuristic for the considered problem has three stages: an initialization stage to generate an initial feasible solution, a local search stage to improve the incumbent solution by searching through different neighborhoods, and a perturbation/shaking stage, wherein the incumbent solution is perturbed to break from a local minimum. In this study, three types of neighborhood searches are employed. Furthermore, two different methods for constructing the initial feasible solution are considered, and multiple variations in the neighborhoods considered are explored in this study. The considered variations and construction methods are evaluated on a total of 128 instances generated with varying vehicle-to-target ratios, distribution for generating the targets, and vehicle-target assignment and are benchmarked against the best-known heuristic for this problem. Two heuristics were finally proposed based on the importance provided to objective value or computation time through extensive computational studies.

Stacking monolayers of transition metal dichalcogenides (TMDs) has led to the discovery of a plethora of new exotic phenomena, resulting from moir\'e pattern formation. Due to the atomic thickness and high surface-to-volume ratio of heterostructures, the interfaces play a crucial role. Fluctuations in the interlayer distance affect interlayer coupling and moir\'e effects. Therefore, to access the intrinsic properties of the TMD stack, it is essential to obtain a clean and uniform interface between the layers. Here, we show that this is achieved by ironing with the tip of an atomic force microscope. This post-stacking procedure dramatically improves the homogeneity of the interfaces, which is reflected in the optical response of the interlayer exciton. We demonstrate that ironing improves the layer coupling, enhancing moir\'e effects and reducing disorder. This is crucial for the investigation of TMD heterostructure physics, which currently suffers from low reproducibility.

We propose Diffusion Model Predictive Control (D-MPC), a novel MPC approach that learns a multi-step action proposal and a multi-step dynamics model, both using diffusion models, and combines them for use in online MPC. On the popular D4RL benchmark, we show performance that is significantly better than existing model-based offline planning methods using MPC and competitive with state-of-the-art (SOTA) model-based and model-free reinforcement learning methods. We additionally illustrate D-MPC's ability to optimize novel reward functions at run time and adapt to novel dynamics, and highlight its advantages compared to existing diffusion-based planning baselines., Comment: Preprint

Multi-modal Large Language Models (MLLMs) exhibit impressive problem-solving abilities in various domains, but their visual comprehension and abstract reasoning skills remain under-evaluated. To this end, we present PolyMATH, a challenging benchmark aimed at evaluating the general cognitive reasoning abilities of MLLMs. PolyMATH comprises 5,000 manually collected high-quality images of cognitive textual and visual challenges across 10 distinct categories, including pattern recognition, spatial reasoning, and relative reasoning. We conducted a comprehensive, and quantitative evaluation of 15 MLLMs using four diverse prompting strategies, including Chain-of-Thought and Step-Back. The best scores achieved on PolyMATH are ~41%, ~36%, and ~27%, obtained by Claude-3.5 Sonnet, GPT-4o and Gemini-1.5 Pro respectively - highlighting the logical and visual complexity of these questions. A further fine-grained error analysis reveals that these models struggle to understand spatial relations and perform drawn-out, high-level reasoning. This is further strengthened by our ablation study estimating MLLM performance when given textual descriptions in place of diagrams. As evidenced by ~4% improvement over textual descriptions as opposed to actual images, we discover that models do not truly comprehend visual diagrams and the spatial information therein, and are thus prone to logical errors. Finally, we evaluate the OpenAI o1 models and find that their performance only matches the human baseline, highlighting the difficulty of the benchmark. The results on PolyMATH highlight the room for improvement in multi-modal reasoning and provide unique insights to guide the development of future MLLMs., Comment: 49 pages, (10 pages paper, 9 pages references, 30 pages appendix)

People's decision-making abilities often fail to improve or may even erode when they rely on AI for decision-support, even when the AI provides informative explanations. We argue this is partly because people intuitively seek contrastive explanations, which clarify the difference between the AI's decision and their own reasoning, while most AI systems offer "unilateral" explanations that justify the AI's decision but do not account for users' thinking. To align human-AI knowledge on decision tasks, we introduce a framework for generating human-centered contrastive explanations that explain the difference between AI's choice and a predicted, likely human choice about the same task. Results from a large-scale experiment (N = 628) demonstrate that contrastive explanations significantly enhance users' independent decision-making skills compared to unilateral explanations, without sacrificing decision accuracy. Amid rising deskilling concerns, our research demonstrates that incorporating human reasoning into AI design can foster human skill development.

We propose the repetition code adapter as a way to perform joint logical Pauli measurements within a quantum low-density parity check (LDPC) codeblock or between separate such codeblocks. This adapter is universal in the sense that it works regardless of the LDPC codes involved and the Paulis being measured. The construction achieves joint logical Pauli measurement of $t$ weight $O(d)$ operators using $O(td\log^2d)$ additional qubits and checks and $O(d)$ time. We also show for some geometrically-local codes in fixed $D\ge2$ dimensions that only $O(td)$ additional qubits and checks are required instead. By extending the adapter in the case $t=2$, we construct a toric code adapter that uses $O(d^2)$ additional qubits and checks to perform targeted logical CNOT gates on arbitrary LDPC codes via Dehn twists. To obtain some of these results, we develop a novel weaker form of graph edge expansion.

Learning from previously collected data via behavioral cloning or offline reinforcement learning (RL) is a powerful recipe for scaling generalist agents by avoiding the need for expensive online learning. Despite strong generalization in some respects, agents are often remarkably brittle to minor visual variations in control-irrelevant factors such as the background or camera viewpoint. In this paper, we present theDeepMind Control Visual Benchmark (DMC-VB), a dataset collected in the DeepMind Control Suite to evaluate the robustness of offline RL agents for solving continuous control tasks from visual input in the presence of visual distractors. In contrast to prior works, our dataset (a) combines locomotion and navigation tasks of varying difficulties, (b) includes static and dynamic visual variations, (c) considers data generated by policies with different skill levels, (d) systematically returns pairs of state and pixel observation, (e) is an order of magnitude larger, and (f) includes tasks with hidden goals. Accompanying our dataset, we propose three benchmarks to evaluate representation learning methods for pretraining, and carry out experiments on several recently proposed methods. First, we find that pretrained representations do not help policy learning on DMC-VB, and we highlight a large representation gap between policies learned on pixel observations and on states. Second, we demonstrate when expert data is limited, policy learning can benefit from representations pretrained on (a) suboptimal data, and (b) tasks with stochastic hidden goals. Our dataset and benchmark code to train and evaluate agents are available at: https://github.com/google-deepmind/dmc_vision_benchmark., Comment: NeurIPS 2024 Datasets and Benchmarks Track. Dataset available at: https://github.com/google-deepmind/dmc_vision_benchmark

This paper addresses the optimization of edge-weighted networks by maximizing algebraic connectivity to enhance network robustness. Motivated by the need for precise robot position estimation in cooperative localization and pose-graph sparsification in Simultaneous Localization and Mapping (SLAM), the algebraic connectivity maximization problem is formulated as a Mixed Integer Semi-Definite Program (MISDP), which is NP-hard. Leveraging spectral graph theoretic methods, specifically Cheeger's inequality, this work introduces novel "Cheeger cuts" to strengthen and efficiently solve medium-scale MISDPs. Further, a new Mixed Integer Linear Program (MILP) is developed for efficiently computing Cheeger cuts, implemented within an outer-approximation algorithm for solving the MISDP. A greedy k-opt heuristic is also presented, producing high-quality solutions that serve as valid lower bounds for Cheeger cuts. Comprehensive numerical analyses demonstrate the efficacy of strengthened cuts via substantial improvements in run times on synthetic and realistic robot localization datasets., Comment: 63rd IEEE Conference on Decision and Control (CDC)

Large Language Models (LLMs), which have gained significant traction in recent years, also function as big structured repositories of data. User personas are a significant and widely utilized method in HCI. This study aims to investigate how LLMs, in their role as data repositories, interpret user personas. Our focus is specifically on personas within the Indian context, seeking to understand how LLMs would interpret such culturally specific personas. To achieve this, we conduct both quantitative and qualitative analyses. This multifaceted approach allows us a primary understanding of the interpretative capabilities of LLMs concerning personas within the Indian context., Comment: At IndiaHCI 2024

In this paper, motion planning for a vehicle moving on a unit sphere with unit speed is considered, wherein the desired terminal location is fixed, but the terminal orientation is free. The motion of the vehicle is modeled to be constrained by a maximum geodesic curvature $U_{max},$ which controls the rate of change of heading of the vehicle such that the maximum heading change occurs when the vehicle travels on a tight circular arc of radius $r = \frac{1}{\sqrt{1 + U_{max}^2}}$. Using Pontryagin's Minimum Principle, the main result of this paper shows that for $r \leq \frac{1}{2}$, the optimal path connecting a given initial configuration and a final location on the sphere belongs to a set of at most seven paths. The candidate paths are of type $CG, CC,$ and degenerate paths of the same, where $C \in \{L, R\}$ denotes a tight left or right turn, respectively, and $G$ denotes a great circular arc., Comment: \c{opyright} 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works

In this paper, motion planning for a Dubins vehicle on a unit sphere to attain a desired final location is considered. The radius of the Dubins path on the sphere is lower bounded by $r$. In a previous study, this problem was addressed, wherein it was shown that the optimal path is of type $CG, CC,$ or a degenerate path of the same for $r \leq \frac{1}{2}.$ Here, $C = L, R$ denotes an arc of a tight left or right turn of minimum turning radius $r,$ and $G$ denotes an arc of a great circle. In this study, the candidate paths for the same problem are generalized to model vehicles with a larger turning radius. In particular, it is shown that the candidate optimal paths are of type $CG, CC,$ or a degenerate path of the same for $r \leq \frac{\sqrt{3}}{2}.$ Noting that at most two $LG$ paths and two $RG$ paths can exist for a given final location, this article further reduces the candidate optimal paths by showing that only one $LG$ and one $RG$ path can be optimal, yielding a total of seven candidate paths for $r \leq \frac{\sqrt{3}}{2}.$ Additional conditions for the optimality of $CC$ paths are also derived in this study.

Quantum circuits constitute Intellectual Property (IP) of the quantum developers and users, which needs to be protected from theft by adversarial agents, e.g., the quantum cloud provider or a rogue adversary present in the cloud. This necessitates the exploration of low-overhead techniques applicable to near-term quantum devices, to trace the quantum circuits/algorithms\textquotesingle{} IP and their output. We present two such lightweight watermarking techniques to prove ownership in the event of an adversary cloning the circuit design. For the first technique, a rotation gate is placed on ancilla qubits combined with other gate(s) at the output of the circuit. For the second method, a set of random gates are inserted in the middle of the circuit followed by its inverse, separated from the circuit by a barrier. These models are combined and applied on benchmark circuits, and the circuit depth, 2-qubit gate count, probability of successful trials (PST), and probabilistic proof of authorship (PPA) are compared against the state-of-the-art. The PST is reduced by a minuscule 0.53\% against the non-watermarked benchmarks and is up to 22.69\% higher compared to existing techniques. The circuit depth has been reduced by up to 27.7\% as against the state-of-the-art. The PPA is astronomically smaller than existing watermarks.

This paper introduces a novel formulation aimed at determining the optimal schedule for recharging a fleet of $n$ heterogeneous robots, with the primary objective of minimizing resource utilization. This study provides a foundational framework applicable to Multi-Robot Mission Planning, particularly in scenarios demanding Long-Duration Autonomy (LDA) or other contexts that necessitate periodic recharging of multiple robots. A novel Integer Linear Programming (ILP) model is proposed to calculate the optimal initial conditions (partial charge) for individual robots, leading to the minimal utilization of charging stations. This formulation was further generalized to maximize the servicing time for robots given adequate charging stations. The efficacy of the proposed formulation is evaluated through a comparative analysis, measuring its performance against the thrift price scheduling algorithm documented in the existing literature. The findings not only validate the effectiveness of the proposed approach but also underscore its potential as a valuable tool in optimizing resource allocation for a range of robotic and engineering applications.

Quantum machine learning (QML) is a rapidly emerging area of research, driven by the capabilities of Noisy Intermediate-Scale Quantum (NISQ) devices. With the progress in the research of QML models, there is a rise in third-party quantum cloud services to cater to the increasing demand for resources. New security concerns surface, specifically regarding the protection of intellectual property (IP) from untrustworthy service providers. One of the most pressing risks is the potential for reverse engineering (RE) by malicious actors who may steal proprietary quantum IPs such as trained parameters and QML architecture, modify them to remove additional watermarks or signatures and re-transpile them for other quantum hardware. Prior work presents a brute force approach to RE the QML parameters which takes exponential time overhead. In this paper, we introduce an autoencoder-based approach to extract the parameters from transpiled QML models deployed on untrusted third-party vendors. We experiment on multi-qubit classifiers and note that they can be reverse-engineered under restricted conditions with a mean error of order 10^-1. The amount of time taken to prepare the dataset and train the model to reverse engineer the QML circuit being of the order 10^3 seconds (which is 10^2x better than the previously reported value for 4-layered 4-qubit classifiers) makes the threat of RE highly potent, underscoring the need for continued development of effective defenses., Comment: 7 pages, 4 figures

Software developers must adapt to keep up with the changing capabilities of platforms so that they can utilize the power of High- Performance Computers (HPC), including exascale systems. OpenMP, a directive-based parallel programming model, allows developers to include directives to existing C, C++, or Fortran code to allow node level parallelism without compromising performance. This paper describes our CI/CD efforts to provide easy evaluation of the support of OpenMP across different compilers using existing testsuites and benchmark suites on HPC platforms. Our main contributions include (1) the set of a Continuous Integration (CI) and Continuous Development (CD) workflow that captures bugs and provides faster feedback to compiler developers, (2) an evaluation of OpenMP (offloading) implementations supported by AMD, HPE, GNU, LLVM, and Intel, and (3) evaluation of the quality of compilers across different heterogeneous HPC platforms. With the comprehensive testing through the CI/CD workflow, we aim to provide a comprehensive understanding of the current state of OpenMP (offloading) support in different compilers and heterogeneous platforms consisting of CPUs and GPUs from NVIDIA, AMD, and Intel.

Quantum machine learning (QML) is a category of algorithms that employ variational quantum circuits (VQCs) to tackle machine learning tasks. Recent discoveries have shown that QML models can effectively generalize from limited training data samples. This capability has sparked increased interest in deploying these models to address practical, real-world challenges, resulting in the emergence of Quantum Machine Learning as a Service (QMLaaS). QMLaaS represents a hybrid model that utilizes both classical and quantum computing resources. Classical computers play a crucial role in this setup, handling initial pre-processing and subsequent post-processing of data to compensate for the current limitations of quantum hardware. Since this is a new area, very little work exists to paint the whole picture of QMLaaS in the context of known security threats in the domain of classical and quantum machine learning. This SoK paper is aimed to bridge this gap by outlining the complete QMLaaS workflow, which encompasses both the training and inference phases and highlighting significant security concerns involving untrusted classical or quantum providers. QML models contain several sensitive assets, such as the model architecture, training/testing data, encoding techniques, and trained parameters. Unauthorized access to these components could compromise the model's integrity and lead to intellectual property (IP) theft. We pinpoint the critical security issues that must be considered to pave the way for a secure QMLaaS deployment., Comment: 9 pages, 3 figures

In this article, two methods of addressing path planning for a Dubins vehicle moving on a sphere are considered, wherein either spherical coordinates or a moving frame are considered to describe the vehicle's motion. The primary contribution of this article is to show the equivalence of these two approaches, which in turn shows that the results known for the moving frame-based description transfer to the model utilizing spherical coordinates.

Medical image segmentation has been traditionally approached by training or fine-tuning the entire model to cater to any new modality or dataset. However, this approach often requires tuning a large number of parameters during training. With the introduction of the Segment Anything Model (SAM) for prompted segmentation of natural images, many efforts have been made towards adapting it efficiently for medical imaging, thus reducing the training time and resources. However, these methods still require expert annotations for every image in the form of point prompts or bounding box prompts during training and inference, making it tedious to employ them in practice. In this paper, we propose an adaptation technique, called S-SAM, that only trains parameters equal to 0.4% of SAM's parameters and at the same time uses simply the label names as prompts for producing precise masks. This not only makes tuning SAM more efficient than the existing adaptation methods but also removes the burden of providing expert prompts. We call this modified version S-SAM and evaluate it on five different modalities including endoscopic images, x-ray, ultrasound, CT, and histology images. Our experiments show that S-SAM outperforms state-of-the-art methods as well as existing SAM adaptation methods while tuning a significantly less number of parameters. We release the code for S-SAM at https://github.com/JayParanjape/SVDSAM., Comment: Accepted in MICCAI 2024

We introduce a novel computational unit, termed PushPull-Conv, in the first layer of a ResNet architecture, inspired by the anti-phase inhibition phenomenon observed in the primary visual cortex. This unit redefines the traditional convolutional layer by implementing a pair of complementary filters: a trainable push kernel and its counterpart, the pull kernel. The push kernel (analogous to traditional convolution) learns to respond to specific stimuli, while the pull kernel reacts to the same stimuli but of opposite contrast. This configuration enhances stimulus selectivity and effectively inhibits response in regions lacking preferred stimuli. This effect is attributed to the push and pull kernels, which produce responses of comparable magnitude in such regions, thereby neutralizing each other. The incorporation of the PushPull-Conv into ResNets significantly increases their robustness to image corruption. Our experiments with benchmark corruption datasets show that the PushPull-Conv can be combined with other data augmentation techniques to further improve model robustness. We set a new robustness benchmark on ResNet50 achieving an $mCE$ of 49.95$\%$ on ImageNet-C when combining PRIME augmentation with PushPull inhibition., Comment: Accepted at ICPR 2024, code available at https://github.com/bgswaroop/pushpull-conv

Multivariate datasets are common in various real-world applications. Recently, copulas have received significant attention for modeling dependencies among random variables. A copula-based information measure is required to quantify the uncertainty inherent in these dependencies. This paper introduces a multivariate variant of the cumulative copula entropy and explores its various properties, including bounds, stochastic orders, and convergence-related results. Additionally, we define a cumulative copula information generating function and derive it for several well-known families of multivariate copulas. A fractional generalization of the multivariate cumulative copula entropy is also introduced and examined. We present a non-parametric estimator of the cumulative copula entropy using empirical beta copula. Furthermore, we propose a new distance measure between two copulas based on the Kullback-Leibler divergence and discuss a goodness-of-fit test based on this measure.

Quantum computing (QC) has the potential to revolutionize fields like machine learning, security, and healthcare. Quantum machine learning (QML) has emerged as a promising area, enhancing learning algorithms using quantum computers. However, QML models are lucrative targets due to their high training costs and extensive training times. The scarcity of quantum resources and long wait times further exacerbate the challenge. Additionally, QML providers may rely on a third-party quantum cloud for hosting the model, exposing the models and training data. As QML-as-a-Service (QMLaaS) becomes more prevalent, reliance on third party quantum clouds can pose a significant threat. This paper shows that adversaries in quantum clouds can use white-box access of the QML model during training to extract the state preparation circuit (containing training data) along with the labels. The extracted training data can be reused for training a clone model or sold for profit. We propose a suite of techniques to prune and fix the incorrect labels. Results show that $\approx$90\% labels can be extracted correctly. The same model trained on the adversarially extracted data achieves approximately $\approx$90\% accuracy, closely matching the accuracy achieved when trained on the original data. To mitigate this threat, we propose masking labels/classes and modifying the cost function for label obfuscation, reducing adversarial label prediction accuracy by $\approx$70\%., Comment: 7 pages

Retrieval augmented generation (RAG) combines the generative abilities of large language models (LLMs) with external knowledge sources to provide more accurate and up-to-date responses. Recent RAG advancements focus on improving retrieval outcomes through iterative LLM refinement or self-critique capabilities acquired through additional instruction tuning of LLMs. In this work, we introduce Speculative RAG - a framework that leverages a larger generalist LM to efficiently verify multiple RAG drafts produced in parallel by a smaller, distilled specialist LM. Each draft is generated from a distinct subset of retrieved documents, offering diverse perspectives on the evidence while reducing input token counts per draft. This approach enhances comprehension of each subset and mitigates potential position bias over long context. Our method accelerates RAG by delegating drafting to the smaller specialist LM, with the larger generalist LM performing a single verification pass over the drafts. Extensive experiments demonstrate that Speculative RAG achieves state-of-the-art performance with reduced latency on TriviaQA, MuSiQue, PubHealth, and ARC-Challenge benchmarks. It notably enhances accuracy by up to 12.97% while reducing latency by 51% compared to conventional RAG systems on PubHealth., Comment: Preprint

Quantum Machine Learning (QML) amalgamates quantum computing paradigms with machine learning models, providing significant prospects for solving complex problems. However, with the expansion of numerous third-party vendors in the Noisy Intermediate-Scale Quantum (NISQ) era of quantum computing, the security of QML models is of prime importance, particularly against reverse engineering, which could expose trained parameters and algorithms of the models. We assume the untrusted quantum cloud provider is an adversary having white-box access to the transpiled user-designed trained QML model during inference. Reverse engineering (RE) to extract the pre-transpiled QML circuit will enable re-transpilation and usage of the model for various hardware with completely different native gate sets and even different qubit technology. Such flexibility may not be obtained from the transpiled circuit which is tied to a particular hardware and qubit technology. The information about the number of parameters, and optimized values can allow further training of the QML model to alter the QML model, tamper with the watermark, and/or embed their own watermark or refine the model for other purposes. In this first effort to investigate the RE of QML circuits, we perform RE and compare the training accuracy of original and reverse-engineered Quantum Neural Networks (QNNs) of various sizes. We note that multi-qubit classifiers can be reverse-engineered under specific conditions with a mean error of order 1e-2 in a reasonable time. We also propose adding dummy fixed parametric gates in the QML models to increase the RE overhead for defense. For instance, adding 2 dummy qubits and 2 layers increases the overhead by ~1.76 times for a classifier with 2 qubits and 3 layers with a performance overhead of less than 9%. We note that RE is a very powerful attack model which warrants further efforts on defenses., Comment: 11 pages, 12 figures

Educating students about academic integrity expectations has been suggested as one of the ways to reduce malpractice in take-home programming assignments. We test this hypothesis using data collected from an artificial intelligence course with 105 participants (N=105) at a university in India. The AI course had two programming assignments. Plagiarism through collusion was quantified using the Measure of Software Similarity (MOSS) tool. Students were educated about what constitutes academic dishonesty and were required to take an honor pledge before the start of the second take-home programming assignment. The two programming assignments were novel and did not have solutions available on the internet. We expected the mean percentage of similar lines of code to be significantly less in the second programming assignment. However, our results show no significant difference in the mean percentage of similar lines of code across the two programming assignments. We also study how the Big-five personality traits affect the propensity for plagiarism in the two take-home assignments. Our results across both assignments show that the extraversion trait of the Big Five personality exhibits a positive association, and the conscientiousness trait exhibits a negative association with plagiarism tendencies. Our result suggests that the policy of educating students about academic integrity will have a limited impact as long as students perceive an opportunity for plagiarism to be present. We explain our results using the Fraud triangle model., Comment: 11 pages, 3 figures, To be published in ACM International Conference on Computing Education Research (ICER) 2024

In recent years, there has been rising interest from both governments and private industry in developing software that is accessible to all, including people with disabilities. However, the computer science (CS) courses that ought to prepare future professionals to develop such accessible software hardly cover topics related to accessibility. While there is growing literature on incorporating accessibility topics in computing education in the West, there is little work on this in the Global South, particularly in India, which has a large number of computing students and software professionals. In this replication report, we present (A) our findings from a replication of surveys used in the US and Switzerland on who teaches accessibility and barriers to teaching accessibility and (B) a qualitative analysis of perceptions of CS faculty in India about digital accessibility and teaching accessibility. Our study corroborates the findings of the earlier surveys: very few CS faculty teach accessibility, and the top barriers they perceive are the same. The qualitative analysis further reveals that the faculty in India need training on accessibility concepts and disabilities sensitization, and exposure to existing and ongoing CS education research and pedagogies. In light of these findings, we present recommendations aimed at addressing these challenges and enhancing the integration of accessibility into computing education., Comment: Accepted to be published in proceedings of the ACM International Conference on Computing Education Research (ICER) 2024

Opinion summarization in e-commerce encapsulates the collective views of numerous users about a product based on their reviews. Typically, a product on an e-commerce platform has thousands of reviews, each review comprising around 10-15 words. While Large Language Models (LLMs) have shown proficiency in summarization tasks, they struggle to handle such a large volume of reviews due to context limitations. To mitigate, we propose a scalable framework called Xl-OpSumm that generates summaries incrementally. However, the existing test set, AMASUM has only 560 reviews per product on average. Due to the lack of a test set with thousands of reviews, we created a new test set called Xl-Flipkart by gathering data from the Flipkart website and generating summaries using GPT-4. Through various automatic evaluations and extensive analysis, we evaluated the framework's efficiency on two datasets, AMASUM and Xl-Flipkart. Experimental results show that our framework, Xl-OpSumm powered by Llama-3-8B-8k, achieves an average ROUGE-1 F1 gain of 4.38% and a ROUGE-L F1 gain of 3.70% over the next best-performing model.

We introduce NATURAL PLAN, a realistic planning benchmark in natural language containing 3 key tasks: Trip Planning, Meeting Planning, and Calendar Scheduling. We focus our evaluation on the planning capabilities of LLMs with full information on the task, by providing outputs from tools such as Google Flights, Google Maps, and Google Calendar as contexts to the models. This eliminates the need for a tool-use environment for evaluating LLMs on Planning. We observe that NATURAL PLAN is a challenging benchmark for state of the art models. For example, in Trip Planning, GPT-4 and Gemini 1.5 Pro could only achieve 31.1% and 34.8% solve rate respectively. We find that model performance drops drastically as the complexity of the problem increases: all models perform below 5% when there are 10 cities, highlighting a significant gap in planning in natural language for SoTA LLMs. We also conduct extensive ablation studies on NATURAL PLAN to further shed light on the (in)effectiveness of approaches such as self-correction, few-shot generalization, and in-context planning with long-contexts on improving LLM planning.

Fitch Cheney's 5-card trick was introduced in 1950. In 2013, Mulcahy invented a 4-card trick in which the cards are allowed to be displayed face down. We suggest our own invention: a 3-card trick in which the cards can be face down and also allowed to be placed both vertically and horizontally. We discuss the theory behind all the tricks and estimate the maximum deck size given the number of chosen cards. We also discuss the cases of hiding several cards and the deck that has duplicates., Comment: 24 pages, 13 tables

Large Language Models (LLMs) excel at various tasks, including solving math word problems (MWPs), but struggle with real-world problems containing irrelevant information. To address this, we propose a prompting framework that generates adversarial variants of MWPs by adding irrelevant variables. We introduce a dataset, PROBLEMATHIC, containing both adversarial and non-adversarial MWPs. Our experiments reveal that LLMs are susceptible to distraction by numerical noise, resulting in an average relative performance drop of ~26% on adversarial MWPs. To mitigate this, we fine-tune LLMs (Llama-2, Mistral) on the adversarial samples from our dataset. Fine-tuning on adversarial training instances improves performance on adversarial MWPs by ~8%, indicating increased robustness to noise and improved ability to identify relevant data for reasoning. Finally, to assess the generalizability of our prompting framework, we introduce GSM-8K-Adv, an adversarial variant of the GSM-8K benchmark. LLMs continue to struggle when faced with adversarial information, reducing performance by up to 6%.

The high expenses imposed by current quantum cloud providers, coupled with the escalating need for quantum resources, may incentivize the emergence of cheaper cloud-based quantum services from potentially untrusted providers. Deploying or hosting quantum models, such as Quantum Neural Networks (QNNs), on these untrusted platforms introduces a myriad of security concerns, with the most critical one being model theft. This vulnerability stems from the cloud provider's full access to these circuits during training and/or inference. In this work, we introduce STIQ, a novel ensemble-based strategy designed to safeguard QNNs against such cloud-based adversaries. Our method innovatively trains two distinct QNNs concurrently, hosting them on same or different platforms, in a manner that each network yields obfuscated outputs rendering the individual QNNs ineffective for adversaries operating within cloud environments. However, when these outputs are combined locally (using an aggregate function), they reveal the correct result. Through extensive experiments across various QNNs and datasets, our technique has proven to effectively masks the accuracy and losses of the individually hosted models by upto $76\%$, albeit at the expense of $\leq 2\times$ increase in the total computational overhead. This trade-off, however, is a small price to pay for the enhanced security and integrity of QNNs in a cloud-based environment prone to untrusted adversaries. We also demonstrated STIQ's practical application by evaluating it on multiple real quantum hardwares, showing that STIQ achieves up to $\approx 70\%$ obfuscation, with combined performance similar to an unobfuscated model.

Quantum Machine Learning (QML) is an accelerating field of study that leverages the principles of quantum computing to enhance and innovate within machine learning methodologies. However, Noisy Intermediate-Scale Quantum (NISQ) computers suffer from noise that corrupts the quantum states of the qubits and affects the training and inferencing accuracy. Furthermore, quantum computers have long access queues. A single execution with a pre-defined number of shots can take hours just to reach the top of the wait queue, which is especially disadvantageous to Quantum Machine Learning (QML) algorithms that are iterative in nature. Many vendors provide access to a suite of quantum hardware with varied qubit technologies, number of qubits, coupling architectures, and noise characteristics. However, present QML algorithms do not use them for the training procedure and often rely on local noiseless/noisy simulators due to cost and training timing overhead on real hardware. Additionally, inferencing is generally performed on reduced datasets with fewer datapoints. Taking these constraints into account, we perform a study to maximize the inferencing performance of QML workloads based on the choice of hardware selection. Specifically, we perform a detailed analysis of quantum classifiers (both training and inference through the lens of hardware queue wait times) on Iris and reduced Digits datasets under noise and varied conditions such as different hardware and coupling maps. We show that using multiple readily available hardware for training rather than relying on a single hardware, especially if it has a long queue depth of pending jobs, can lead to a performance impact of only 3-4% while providing up to 45X reduction in training wait time., Comment: 8 pages, 8 figures/plots, 6 tables

In recent years, various large foundation models have been proposed for image segmentation. There models are often trained on large amounts of data corresponding to general computer vision tasks. Hence, these models do not perform well on medical data. There have been some attempts in the literature to perform parameter-efficient finetuning of such foundation models for medical image segmentation. However, these approaches assume that all the parameters of the model are available for adaptation. But, in many cases, these models are released as APIs or blackboxes, with no or limited access to the model parameters and data. In addition, finetuning methods also require a significant amount of compute, which may not be available for the downstream task. At the same time, medical data can't be shared with third-party agents for finetuning due to privacy reasons. To tackle these challenges, we pioneer a blackbox adaptation technique for prompted medical image segmentation, called BAPS. BAPS has two components - (i) An Image-Prompt decoder (IP decoder) module that generates visual prompts given an image and a prompt, and (ii) A Zero Order Optimization (ZOO) Method, called SPSA-GC that is used to update the IP decoder without the need for backpropagating through the foundation model. Thus, our method does not require any knowledge about the foundation model's weights or gradients. We test BAPS on four different modalities and show that our method can improve the original model's performance by around 4%., Comment: Accepted early at MICCAI 2024

In this paper, we investigate the impact of bit flip errors in FPGA memories in control electronics on quantum computing systems. FPGA memories are integral in storing the amplitude and phase information pulse envelopes, which are essential for generating quantum gate pulses. However, these memories can incur faults due to physical and environmental stressors such as electromagnetic interference, power fluctuations, and temperature variations and adversarial fault injections, potentially leading to errors in quantum gate operations. To understand how these faults affect quantum computations, we conducted a series of experiments to introduce bit flips into the amplitude (both real and imaginary components) and phase values of quantum pulses using IBM's simulated quantum environments, FakeValencia, FakeManila, and FakeLima. Our findings reveal that bit flips in the exponent and initial mantissa bits of the real amplitude cause substantial deviations in quantum gate operations, with TVD increases as high as ~200%. Interestingly, the remaining bits exhibited natural tolerance to errors. We proposed a 3-bit repetition error correction code, which effectively reduced the TVD increases to below 40% without incurring any memory overhead. Due to reuse of less significant bits for error correction, the proposed approach introduces maximum of 5-7% extra TVD in nominal cases. However, this can be avoided by sacrificing memory area for implementing the repetition code., Comment: 9 pages, 9 figures, conference

Quantum computing (QC) is poised to revolutionize problem solving across various fields, with research suggesting that systems with over 50 qubits may achieve quantum advantage surpassing supercomputers in certain optimization tasks. As the hardware size of Noisy Intermediate Scale Quantum (NISQ) computers continues to grow, Multi tenant computing (MTC) has emerged as a viable approach to enhance hardware utilization by allowing shared resource access across multiple quantum programs. However, MTC can also bring challenges and security concerns. This paper focuses on optimizing quantum hardware utilization in shared environments by implementing multi programming strategies that not only enhance hardware utilization but also effectively manage associated risks like crosstalk and fault injection. We propose a novel partitioning and allocation method called Community Based Dynamic Allocation Partitioning (COMDAP) and Secure COMDAP to refine and secure multi programming capabilities in quantum systems. COMDAP ensures equitable and efficient resource distribution, addresses the issues of suboptimal partitioning, and significantly improves hardware utilization. We report a 23 percent average improvement in hardware utilization rate compared to existing greedy heuristics, with rates averaging 92 percent. COMDAP introduces an average increase of approximately 0.05X in delta CX, alongside a 3.5 percent average reduction in PST across benchmarks., Comment: 10 pages, 10 figures

Ensuring safety is the most important factor in connected and autonomous vehicles, especially in emergency braking situations. As such, assessing the safety benefits of one information topology over other is a necessary step towards evaluating and ensuring safety. In this paper, we compare the safety benefits of a cooperative adaptive cruise control which utilizes information from one predecessor vehicle (CACC) with the one that utilizes information from multiple predecessors (CACC+) for the maintenance of spacing under an emergency braking scenario. A constant time headway policy is employed for maintenance of spacing (that includes a desired standstill spacing distance and a velocity dependent spacing distance) between the vehicles in the platoon. The considered emergency braking scenario consists of braking of the leader vehicle of the platoon at its maximum deceleration and that of the following vehicles to maintain the spacing as per CACC or CACC+. By focusing on the standstill spacing distance and utilizing Monte Carlo simulations, we assess the safety benefits of CACC+ over CACC by utilizing the following safety metrics: (1) probability of collision, (2) expected number of collisions, and (3) severity of collision (defined as the relative velocity of the two vehicles at impact). We present and provide discussion of these results.

Quantum Generative Adversarial Networks (qGANs) are at the forefront of image-generating quantum machine learning models. To accommodate the growing demand for Noisy Intermediate-Scale Quantum (NISQ) devices to train and infer quantum machine learning models, the number of third-party vendors offering quantum hardware as a service is expected to rise. This expansion introduces the risk of untrusted vendors potentially stealing proprietary information from the quantum machine learning models. To address this concern we propose a novel watermarking technique that exploits the noise signature embedded during the training phase of qGANs as a non-invasive watermark. The watermark is identifiable in the images generated by the qGAN allowing us to trace the specific quantum hardware used during training hence providing strong proof of ownership. To further enhance the security robustness, we propose the training of qGANs on a sequence of multiple quantum hardware, embedding a complex watermark comprising the noise signatures of all the training hardware that is difficult for adversaries to replicate. We also develop a machine learning classifier to extract this watermark robustly, thereby identifying the training hardware (or the suite of hardware) from the images generated by the qGAN validating the authenticity of the model. We note that the watermark signature is robust against inferencing on hardware different than the hardware that was used for training. We obtain watermark extraction accuracy of 100% and ~90% for training the qGAN on individual and multiple quantum hardware setups (and inferencing on different hardware), respectively. Since parameter evolution during training is strongly modulated by quantum noise, the proposed watermark can be extended to other quantum machine learning models as well., Comment: 11 pages, 10 figures

Quantum error correction (QEC) plays a crucial role in correcting noise and paving the way for fault-tolerant quantum computing. This field has seen significant advancements, with new quantum error correction codes emerging regularly to address errors effectively. Among these, topological codes, particularly surface codes, stand out for their low error thresholds and feasibility for implementation in large-scale quantum computers. However, these codes are restricted to encoding a single qubit. Lattice surgery is crucial for enabling interactions among multiple encoded qubits or between the lattices of a surface code, ensuring that its sophisticated error-correcting features are maintained without significantly increasing the operational overhead. Lattice surgery is pivotal for scaling QECCs across more extensive quantum systems. Despite its critical importance, comprehending lattice surgery is challenging due to its inherent complexity, demanding a deep understanding of intricate quantum physics and mathematical concepts. This paper endeavors to demystify lattice surgery, making it accessible to those without a profound background in quantum physics or mathematics. This work explores surface codes, introduces the basics of lattice surgery, and demonstrates its application in building quantum gates and emulating multi-qubit circuits., Comment: 10 pages, 13 figures, 1 table

In this paper, we investigate the selection of time headway to ensure robust string stability in connected and autonomous vehicle platoons in the presence of signal noise in Vehicle-to-Vehicle (V2V) communication. In particular, we consider the effect of noise in communicated vehicle acceleration from the predecessor vehicle to the follower vehicle on the selection of the time headway in predecessor-follower type vehicle platooning with a Constant Time Headway Policy (CTHP). Employing a CTHP based control law for each vehicle that utilizes on-board sensors for measurement of position and velocity of the predecessor vehicle and wireless communication network for obtaining the acceleration of the predecessor vehicle, we investigate how time headway is affected by communicated signal noise. We derive constraints on the CTHP controller gains for predecessor acceleration, velocity error and spacing error and a lower bound on the time headway which will ensure robust string stability of the platoon against signal noise. We provide comparative numerical simulations on an example to illustrate the main result.

In this paper, we investigate the effect of signal delay in communicated information in connected and autonomous vehicles. In particular, we relate this delay's effect on the selection of the time headway in predecessor-follower type vehicle platooning with a constant time headway policy (CTHP). We employ a CTHP control law for each vehicle in the platoon by considering two cases: cooperative adaptive cruise control (CACC) strategy where information from only one predecessor vehicle is employed and CACC+ where information from multiple predecessor vehicles is employed. We investigate how the lower bound on the time headway is affected by signal transmission delay due to wireless communication. We provide a systematic approach to the derivation of the lower bound of the time headway and selection of the appropriate CTHP controller gains for predecessor acceleration, velocity error and spacing error which will ensure robust string stability of the platoon under the presence of signal delay. We corroborate the main result with numerical simulations.

IL-3 is a hemopoietic growth factor that usually targets blood cell precursors; IL-3R is a cytokine receptor that binds to IL-3. However, IL-3 takes on a different role in the context of glial cells in the nervous system, where studies show that the protein IL-3 protects against Alzheimer's disease by activating microglia at their IL-3R receptors, causing the microglia to clear out the tangles caused by the build-up of misfolded Tau proteins. In this study, we seek to ascertain what role the secondary structure of IL-3 plays in its binding with the receptor. The motivation behind this study is to learn more about the mechanism and identify possible drugs that might be able to activate it, in hopes of inhibiting the spread of Alzheimer's Disease. From a preliminary analysis of complexes containing IL-3 and IL-3R, we hypothesized that the binding is largely due to the interactions of three alpha helix structures stretching towards the active site on the receptor. The original Il-3 protein serves as the control in this experiment; the other proteins being tested are generated through several types of computational de novo protein design, where machine learning allows for the production of entirely novel structures. The efficacy of the generated proteins is assessed through docking simulations with the IL-3R receptor, and the binding poses are also qualitatively examined to gain insight into the function of the binding. From the docking data and poses, the most successful proteins were those with similar secondary structure to IL-3., Comment: 9 pages total

In North America, there are many diverse species of native bees crucial for the environment, who are the primary pollinators of most native floral species. The Californian agriculture industry imports European honeybees (Apis Mellifera) primarily for pollinating almonds. Unfortunately, this has resulted in the unintended consequence of disrupting the native ecosystem and threatening many native bee species as they are outcompeted for food. Our first step for protecting the native species is identification with the use of a Convolutional Neural Network (CNN) to differentiate common native bee species from invasive ones. Removing invasive colonies efficiently without harming native species is difficult as pesticides cause myriad diseases in native species. Our approach seeks to prevent the formation of new queens, causing the colony's collapse. Workers secrete royal jelly, a substance that causes fertility and longevity; it is fed to future honeybee queens. Targeting the production of this substance is safe as no native species use it; small organic molecules (ligands) prevent the proteins Apisimin and MRJP1 from combining and producing an oligomer used to form the substance. Ideal ligands bind to only one of these proteins preventing them from joining together: they have a high affinity for one receptor and a significantly lower affinity for the other. We optimized the CNN to provide a framework for creating Machine Learning models that excel at differentiating between subspecies of insects by measuring the effects of image alteration and class grouping on model performance. The CNN is able to achieve an accuracy of 82% in differentiating between invasive and native bee species; 3 ligands have been identified as effective. Our new approach offers a promising solution to curb the spread of invasive bees within California through an identification and neutralization method., Comment: 13 pages total

The phase-field method has become a useful tool for the simulation of classical metallurgical phase transformations as well as other phenomena related to materials science. The thermodynamic consistency that forms the basis of these formulations lends to its strong predictive capabilities and utility. However, a strong impediment to the usage of the method for typical applied problems of industrial and academic relevance is the significant overhead with regard to the code development and know-how required for quantitative model formulations. In this paper, we report the development of an open-source phase-field software stack that contains generic formulations for the simulation of multi-phase and multi-component phase transformations. The solvers incorporate thermodynamic coupling that allows the realization of simulations with real alloys in scenarios directly relevant to the materials industry. Further, the solvers utilize parallelization strategies using either multiple CPUs or GPUs to provide cross-platform portability and usability on available supercomputing machines. Finally, the solver stack also contains a graphical user interface to gradually introduce the usage of the software. The user interface also provides a collection of post-processing tools that allow the estimation of useful metrics related to microstructural evolution.