Search

Multi-valued Byzantine agreement (MVBA) protocols are essential for atomic broadcast and fault-tolerant state machine replication in asynchronous networks. Despite advances, challenges persist in optimizing these protocols for communication and computation efficiency. This paper presents a committee-based MVBA protocol (cMVBA), a novel approach that achieves agreement without extra communication rounds by analyzing message patterns in asynchronous networks with probability 1., Comment: arXiv admin note: text overlap with arXiv:2406.03739

The challenge of mapping AI architectures to GPU hardware is creating a critical bottleneck in AI progress. Despite substantial efforts, hand-written custom kernels fail to meet their theoretical performance thresholds, even on well-established operations like linear attention. The diverse hardware capabilities of GPUs might suggest that we need a wide variety of techniques to achieve high performance. However, our work explores whether a small number of key abstractions can drastically simplify the process. We present ThunderKittens (TK), a framework for writing performant AI kernels while remaining easy to use and maintain. Our abstractions map to the three levels of the GPU hierarchy: (1) at the warp-level, we provide 16x16 matrix tiles as basic data structures and PyTorch-like parallel compute operations over tiles, (2) at the thread-block level, we provide a template for overlapping asynchronous operations across parallel warps, and (3) at the grid-level, we provide support to help hide the block launch and tear-down, and memory costs. We show the value of TK by providing kernels that match or outperform prior kernels for a range of AI operations. We match CuBLAS and FlashAttention-3 on GEMM and attention inference performance and outperform the strongest baselines by $10-40\%$ on attention backwards, $8\times$ on state space models, and $14\times$ on linear attention.

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

Offloading computationally expensive algorithms to the edge or even cloud offers an attractive option to tackle limitations regarding on-board computational and energy resources of robotic systems. In cloud-native applications deployed with the container management system Kubernetes (K8s), one key problem is ensuring resilience against various types of failures. However, complex robotic systems interacting with the physical world pose a very specific set of challenges and requirements that are not yet covered by failure mitigation approaches from the cloud-native domain. In this paper, we therefore propose a novel approach for robotic system monitoring and stateful, reactive failure mitigation for distributed robotic systems deployed using Kubernetes (K8s) and the Robot Operating System (ROS2). By employing the generic substrate of Behaviour Trees, our approach can be applied to any robotic workload and supports arbitrarily complex monitoring and failure mitigation strategies. We demonstrate the effectiveness and application-agnosticism of our approach on two example applications, namely Autonomous Mobile Robot (AMR) navigation and robotic manipulation in a simulated environment.

Mobile systems will have to support multiple AI-based applications, each leveraging heterogeneous data sources through DNN architectures collaboratively executed within the network. To minimize the cost of the AI inference task subject to requirements on latency, quality, and - crucially - reliability of the inference process, it is vital to optimize (i) the set of sensors/data sources and (ii) the DNN architecture, (iii) the network nodes executing sections of the DNN, and (iv) the resources to use. To this end, we leverage dynamic gated neural networks with branches, and propose a novel algorithmic strategy called Quantile-constrained Inference (QIC), based upon quantile-Constrained policy optimization. QIC makes joint, high-quality, swift decisions on all the above aspects of the system, with the aim to minimize inference energy cost. We remark that this is the first contribution connecting gated dynamic DNNs with infrastructure-level decision making. We evaluate QIC using a dynamic gated DNN with stems and branches for optimal sensor fusion and inference, trained on the RADIATE dataset offering Radar, LiDAR, and Camera data, and real-world wireless measurements. Our results confirm that QIC matches the optimum and outperforms its alternatives by over 80%.

The rise of digital misinformation has heightened interest in using multilingual Large Language Models (LLMs) for fact-checking. This study systematically evaluates translation bias and the effectiveness of LLMs for cross-lingual claim verification across 15 languages from five language families: Romance, Slavic, Turkic, Indo-Aryan, and Kartvelian. Using the XFACT dataset to assess their impact on accuracy and bias, we investigate two distinct translation methods: pre-translation and self-translation. We use mBERT's performance on the English dataset as a baseline to compare language-specific accuracies. Our findings reveal that low-resource languages exhibit significantly lower accuracy in direct inference due to underrepresentation in the training data. Furthermore, larger models demonstrate superior performance in self-translation, improving translation accuracy and reducing bias. These results highlight the need for balanced multilingual training, especially in low-resource languages, to promote equitable access to reliable fact-checking tools and minimize the risk of spreading misinformation in different linguistic contexts., Comment: Accepted to ATTRIB @ NeurIPS 2024

Recent works show we can linearize large language models (LLMs) -- swapping the quadratic attentions of popular Transformer-based LLMs with subquadratic analogs, such as linear attention -- avoiding the expensive pretraining costs. However, linearizing LLMs often significantly degrades model quality, still requires training over billions of tokens, and remains limited to smaller 1.3B to 7B LLMs. We thus propose Low-rank Linear Conversion via Attention Transfer (LoLCATs), a simple two-step method that improves LLM linearizing quality with orders of magnitudes less memory and compute. We base these steps on two findings. First, we can replace an LLM's softmax attentions with closely-approximating linear attentions, simply by training the linear attentions to match their softmax counterparts with an output MSE loss ("attention transfer"). Then, this enables adjusting for approximation errors and recovering LLM quality simply with low-rank adaptation (LoRA). LoLCATs significantly improves linearizing quality, training efficiency, and scalability. We significantly reduce the linearizing quality gap and produce state-of-the-art subquadratic LLMs from Llama 3 8B and Mistral 7B v0.1, leading to 20+ points of improvement on 5-shot MMLU. Furthermore, LoLCATs does so with only 0.2% of past methods' model parameters and 0.4% of their training tokens. Finally, we apply LoLCATs to create the first linearized 70B and 405B LLMs (50x larger than prior work). When compared with prior approaches under the same compute budgets, LoLCATs significantly improves linearizing quality, closing the gap between linearized and original Llama 3.1 70B and 405B LLMs by 77.8% and 78.1% on 5-shot MMLU., Comment: 47 pages, 20 figures, 18 tables, preprint

In this article, we will characterize Weyl multipliers for the pair $(L^1(G \times \widehat{G}; A), L^p(G \times \widehat{G};A))$, for $1 \leq p< \infty$, under the assumption that $A$ is a complex Banach algebra with a bounded approximate identity.

Low-Rank Adaptation (LoRA) is a popular technique for efficient fine-tuning of foundation models. However, applying LoRA in federated learning environments, where data is distributed across multiple clients, presents unique challenges. Existing methods rely on traditional federated averaging of LoRA adapters, resulting in inexact updates. To address this, we propose Federated Exact LoRA, or FedEx-LoRA, which adds a residual error term to the pretrained frozen weight matrix. Our approach achieves exact updates with minimal computational and communication overhead, preserving LoRA's efficiency. We evaluate the method on various Natural Language Understanding (NLU) and Natural Language Generation (NLG) tasks, showing consistent performance gains over state-of-the-art methods across multiple settings. Through extensive analysis, we quantify that the deviations in updates from the ideal solution are significant, highlighting the need for exact aggregation. Our method's simplicity, efficiency, and broad applicability position it as a promising solution for accurate and effective federated fine-tuning of foundation models., Comment: RS and KP contributed equally to this work: 18 Pages, 9 Figures, and 8 Tables. Another version of the paper accepted at NeurIPS 2024 Workshop on Fine-Tuning in Modern Machine Learning: Principles and Scalability

Organic optoelectronic materials are a promising avenue for next-generation electronic devices due to their solution processability, mechanical flexibility, and tunable electronic properties. In particular, near-infrared (NIR) sensitive molecules have unique applications in night-vision equipment and biomedical imaging. Molecular engineering has played a crucial role in developing non-fullerene acceptors (NFAs) such as the Y-series molecules, which have significantly improved the power conversion efficiency (PCE) of solar cells and enhanced spectral coverage in the NIR region. However, systematically designing molecules with targeted optoelectronic properties while ensuring synthetic accessibility remains a challenge. To address this, we leverage structural priors from domain-focused, patent-mined datasets of organic electronic molecules using a symmetry-aware fragment decomposition algorithm and a fragment-constrained Monte Carlo Tree Search (MCTS) generator. Our approach generates candidates that retain symmetry constraints from the patent dataset, while also exhibiting red-shifted absorption, as validated by TD-DFT calculations.

Edge computing allows for reduced latency and operational costs compared to centralized cloud systems. In this context, serverless functions are emerging as a lightweight and effective paradigm for managing computational tasks on edge infrastructures. However, the placement of such functions in constrained edge nodes remains an open challenge. On one hand, it is key to minimize network delays and optimize resource consumption; on the other hand, decisions must be made in a timely manner due to the highly dynamic nature of edge environments. In this paper, we propose POSEIDON, a solution based on Deep Reinforcement Learning for the efficient placement of functions at the edge. POSEIDON leverages Proximal Policy Optimization (PPO) to place functions across a distributed network of nodes under highly dynamic workloads. A comprehensive empirical evaluation demonstrates that POSEIDON significantly reduces execution time, network delay, and resource consumption compared to state-of-the-art methods., Comment: This paper is accepted at ICSOC'24 (International Conference on Service-Oriented Computing)

The spatio-temporal complexity of video data presents significant challenges in tasks such as compression, generation, and inpainting. We present four key contributions to address the challenges of spatiotemporal video processing. First, we introduce the 3D Mobile Inverted Vector-Quantization Variational Autoencoder (3D-MBQ-VAE), which combines Variational Autoencoders (VAEs) with masked token modeling to enhance spatiotemporal video compression. The model achieves superior temporal consistency and state-of-the-art (SOTA) reconstruction quality by employing a novel training strategy with full frame masking. Second, we present MotionAura, a text-to-video generation framework that utilizes vector-quantized diffusion models to discretize the latent space and capture complex motion dynamics, producing temporally coherent videos aligned with text prompts. Third, we propose a spectral transformer-based denoising network that processes video data in the frequency domain using the Fourier Transform. This method effectively captures global context and long-range dependencies for high-quality video generation and denoising. Lastly, we introduce a downstream task of Sketch Guided Video Inpainting. This task leverages Low-Rank Adaptation (LoRA) for parameter-efficient fine-tuning. Our models achieve SOTA performance on a range of benchmarks. Our work offers robust frameworks for spatiotemporal modeling and user-driven video content manipulation. We will release the code, datasets, and models in open-source., Comment: Under submission at a conference

STImulated Raman Adiabatic Passage (STIRAP) is a powerful technique for robust state transfer capabilities in quantum systems. This method, however encounters challenges for its implementation as a gate in qubit-subspace due to its sensitivity to initial states. By incorporating single-photon detuning into the protocol, the sensitivity to the initial state can effectively be mitigated, enabling STIRAP to operate as a gate. In this study, we experimentally demonstrate the implementation of robust $\pi$ and $\pi$/2 rotations in a dual-rail qubit formed by two strongly coupled fixed-frequency transmon qubits. We achieve state preparation fidelity in excess of 0.98 using such rotations. Our analysis reveals these gates exhibit significant resilience to errors. Furthermore, our numerical calculations confirm that these gates can achieve fidelity levels in excess of 0.999. This work suggest a way for realizing quantum gates which are robust against minor drifts in pulse or system parameters.

The Byzantine Agreement (BA) problem is a fundamental challenge in distributed systems, focusing on achieving reaching an agreement among parties, some of which may behave maliciously. With the rise of cryptocurrencies, there has been significant interest in developing atomic broadcast protocols, which facilitate agreement on a subset of parties' requests. However, these protocols often come with high communication complexity ($O(ln^2 + \lambda n^3 \log n)$, where $l$ is the bit length of the input, $n$ is the number of parties, and $\lambda$ represents the security parameter bit length). This can lead to inefficiency, especially when the requests across parties exhibit little variation, resulting in unnecessary resource consumption. In this paper, we introduce Slim-ABC, a novel atomic broadcast protocol that eliminates the $O(ln^2 + \lambda n^3 \log n)$ term associated with traditional atomic broadcast protocols. While Slim-ABC reduces the number of accepted requests, it significantly mitigates resource wastage, making it more efficient. The protocol leverages the asynchronous common subset and provable-broadcast mechanisms to achieve a communication complexity of $O(ln^2 + \lambda n^2)$. Despite the trade-off in accepted requests, Slim-ABC maintains robust security by allowing only a fraction ($f+1$) of parties to broadcast requests. We present an extensive efficiency analysis of Slim-ABC, evaluating its performance across key metrics such as message complexity, communication complexity, and time complexity. Additionally, we provide a rigorous security analysis, demonstrating that Slim-ABC satisfies the \textit{agreement}, \textit{validity}, and \textit{totality} properties of the asynchronous common subset protocol., Comment: 21 pages

One of the prominent issues stifling the current generation of large language models is their limited context length. Recent proprietary models such as GPT-4 and Claude 2 have introduced longer context lengths, 8k/32k and 100k, respectively; however, despite the efforts in the community, most common models, such as LLama-2, have a context length of 4k or less. Unlimiformer (Bertsch et al., 2023) is a recently popular vector-retrieval augmentation method that offloads cross-attention computations to a kNN index. However, its main limitation is incompatibility with decoder-only transformers out of the box. In this work, we explore practical considerations of adapting Unlimiformer to decoder-only transformers and introduce a series of modifications to overcome this limitation. Moreover, we expand the original experimental setup on summarization to include a new task (i.e., free-form Q&A) and an instruction-tuned model (i.e., a custom 6.7B GPT model). Our results showcase the effectiveness of these modifications on summarization, performing on par with a model with 2x the context length. Moreover, we discuss limitations and future directions for free-form Q&A and instruction-tuned models., Comment: 8 pages, 6 figures

Model merging, such as model souping, is the practice of combining different models with the same architecture together without further training. In this work, we present a model merging methodology that addresses the difficulty of fine-tuning Large Language Models (LLMs) for target tasks in non-English languages, where task-specific data is often unavailable. We focus on mathematical reasoning and without in-language math data, facilitate cross-lingual transfer by composing language and math capabilities. Starting from the same pretrained model, we fine-tune separate "experts" on math instruction data in English and on generic instruction data in the target language. We then replace the top and bottom transformer layers of the math expert directly with layers from the language expert, which consequently enhances math performance in the target language. The resulting merged models outperform the individual experts and other merging methods on the math benchmark, MGSM, by 10% across four major languages where math instruction data is scarce. In addition, this layer swapping is simple, inexpensive, and intuitive, as it is based on an interpretative analysis of the most important parameter changes during the fine-tuning of each expert. The ability to successfully re-compose LLMs for cross-lingual transfer in this manner opens up future possibilities to combine model expertise, create modular solutions, and transfer reasoning capabilities across languages all post hoc., Comment: 11 main pages, 23 pages total, 9 figures, 5 tables

Deep multimodal learning has shown remarkable success by leveraging contrastive learning to capture explicit one-to-one relations across modalities. However, real-world data often exhibits shared relations beyond simple pairwise associations. We propose M3CoL, a Multimodal Mixup Contrastive Learning approach to capture nuanced shared relations inherent in multimodal data. Our key contribution is a Mixup-based contrastive loss that learns robust representations by aligning mixed samples from one modality with their corresponding samples from other modalities thereby capturing shared relations between them. For multimodal classification tasks, we introduce a framework that integrates a fusion module with unimodal prediction modules for auxiliary supervision during training, complemented by our proposed Mixup-based contrastive loss. Through extensive experiments on diverse datasets (N24News, ROSMAP, BRCA, and Food-101), we demonstrate that M3CoL effectively captures shared multimodal relations and generalizes across domains. It outperforms state-of-the-art methods on N24News, ROSMAP, and BRCA, while achieving comparable performance on Food-101. Our work highlights the significance of learning shared relations for robust multimodal learning, opening up promising avenues for future research., Comment: RK and RS contributed equally to this work, 20 Pages, 8 Figures, 9 Tables. Another version of the paper accepted at NeurIPS 2024 Workshop on Unifying Representations in Neural Models (UniReps)

Positron Emission Tomography (PET) /Computed Tomography (CT) is crucial for diagnosing, managing, and planning treatment for various cancers. Developing reliable deep learning models for the segmentation of tumor lesions in PET/CT scans in a multi-tracer multicenter environment, is a critical area of research. Different tracers, such as Fluorodeoxyglucose (FDG) and Prostate-Specific Membrane Antigen (PSMA), have distinct physiological uptake patterns and data from different centers often vary in terms of acquisition protocols, scanner types, and patient populations. Because of this variability, it becomes more difficult to design reliable segmentation algorithms and generalization techniques due to variations in image quality and lesion detectability. To address this challenge, We trained a 3D Residual encoder U-Net within the no new U-Net framework, aiming to generalize the performance of automatic lesion segmentation of whole body PET/CT scans, across different tracers and clinical sites. Further, We explored several preprocessing techniques and ultimately settled on using the Total Segmentator to crop our training data. Additionally, we applied resampling during this process. During inference, we leveraged test-time augmentations and other post-processing techniques to enhance tumor lesion segmentation. Our team currently hold the top position in the Auto-PET III challenge and outperformed the challenge baseline model in the preliminary test set with Dice score of 0.9627.

Data poisoning and leakage risks impede the massive deployment of federated learning in the real world. This chapter reveals the truths and pitfalls of understanding two dominating threats: {\em training data privacy intrusion} and {\em training data poisoning}. We first investigate training data privacy threat and present our observations on when and how training data may be leaked during the course of federated training. One promising defense strategy is to perturb the raw gradient update by adding some controlled randomized noise prior to sharing during each round of federated learning. We discuss the importance of determining the proper amount of randomized noise and the proper location to add such noise for effective mitigation of gradient leakage threats against training data privacy. Then we will review and compare different training data poisoning threats and analyze why and when such data poisoning induced model Trojan attacks may lead to detrimental damage on the performance of the global model. We will categorize and compare representative poisoning attacks and the effectiveness of their mitigation techniques, delivering an in-depth understanding of the negative impact of data poisoning. Finally, we demonstrate the potential of dynamic model perturbation in simultaneously ensuring privacy protection, poisoning resilience, and model performance. The chapter concludes with a discussion on additional risk factors in federated learning, including the negative impact of skewness, data and algorithmic biases, as well as misinformation in training data. Powered by empirical evidence, our analytical study offers some transformative insights into effective privacy protection and security assurance strategies in attack-resilient federated learning., Comment: Chapter of Handbook of Trustworthy Federated Learning

Chain-of-thought (CoT) via prompting is the de facto method for eliciting reasoning capabilities from large language models (LLMs). But for what kinds of tasks is this extra ``thinking'' really helpful? To analyze this, we conducted a quantitative meta-analysis covering over 100 papers using CoT and ran our own evaluations of 20 datasets across 14 models. Our results show that CoT gives strong performance benefits primarily on tasks involving math or logic, with much smaller gains on other types of tasks. On MMLU, directly generating the answer without CoT leads to almost identical accuracy as CoT unless the question or model's response contains an equals sign, indicating symbolic operations and reasoning. Following this finding, we analyze the behavior of CoT on these problems by separating planning and execution and comparing against tool-augmented LLMs. Much of CoT's gain comes from improving symbolic execution, but it underperforms relative to using a symbolic solver. Our results indicate that CoT can be applied selectively, maintaining performance while saving inference costs. Furthermore, they suggest a need to move beyond prompt-based CoT to new paradigms that better leverage intermediate computation across the whole range of LLM applications., Comment: Swapped column names for Table 7 and 8 in the appendix. Fixed the prompt for SocialIQA; results in figures and tables are updated (no major differences, but the prompt is now correct)

The charge neutrality point of bilayer graphene, denoted as {\nu} = 0 state, manifests competing phases marked by spontaneously broken isospin (spin/valley/layer) symmetries under external magnetic and electric fields. However, due to their electrically insulating nature, identifying these phases through electrical conductance measurements remains challenging. A recent theoretical proposal introduces a novel approach, employing thermal transport measurements to detect these competing phases. Here, we experimentally explore the bulk thermal transport of the {\nu} = 0 state in bilayer graphene to investigate its ground states and collective excitations associated with isospin. While the theory anticipates a finite thermal conductance in the {\nu} = 0 state, our findings unveil an absence of detectable thermal conductance. Through variations in the external electric field and temperature-dependent measurements, our results suggest towards gapped collective excitations at {\nu} = 0 state. Our findings underscore the necessity for further investigations into the nature of {\nu} = 0.

Given the widespread dissemination of misinformation on social media, implementing fact-checking mechanisms for online claims is essential. Manually verifying every claim is very challenging, underscoring the need for an automated fact-checking system. This paper presents our system designed to address this issue. We utilize the Averitec dataset (Schlichtkrull et al., 2023) to assess the performance of our fact-checking system. In addition to veracity prediction, our system provides supporting evidence, which is extracted from the dataset. We develop a Retrieve and Generate (RAG) pipeline to extract relevant evidence sentences from a knowledge base, which are then inputted along with the claim into a large language model (LLM) for classification. We also evaluate the few-shot In-Context Learning (ICL) capabilities of multiple LLMs. Our system achieves an 'Averitec' score of 0.33, which is a 22% absolute improvement over the baseline. Our Code is publicly available on https://github.com/ronit-singhal/evidence-backed-fact-checking-using-rag-and-few-shot-in-context-learning-with-llms., Comment: Accepted in The Seventh FEVER Workshop at EMNLP 2024

We show that the most well-known and fundamental building blocks of DP implementations -- sum, mean, count (and many other linear queries) -- can be released with substantially reduced noise for the same privacy guarantee. We achieve this by projecting individual data with worst-case sensitivity $R$ onto a simplex where all data now has a constant norm $R$. In this simplex, additional ``free'' queries can be run that are already covered by the privacy-loss of the original budgeted query, and which algebraically give additional estimates of counts or sums, and can be combined for lower final noise., Comment: TPDP24

Synthetic infrared (IR) scene and target generation is an important computer vision problem as it allows the generation of realistic IR images and targets for training and testing of various applications, such as remote sensing, surveillance, and target recognition. It also helps reduce the cost and risk associated with collecting real-world IR data. This survey paper aims to provide a comprehensive overview of the conventional mathematical modelling-based methods and deep learning-based methods used for generating synthetic IR scenes and targets. The paper discusses the importance of synthetic IR scene and target generation and briefly covers the mathematics of blackbody and grey body radiations, as well as IR image-capturing methods. The potential use cases of synthetic IR scenes and target generation are also described, highlighting the significance of these techniques in various fields. Additionally, the paper explores possible new ways of developing new techniques to enhance the efficiency and effectiveness of synthetic IR scenes and target generation while highlighting the need for further research to advance this field., Comment: Submitted in Journal of Infrared Physics & Technology

The application of Large Language Models (LLMs) in healthcare is expanding rapidly, with one potential use case being the translation of formal medical reports into patient-legible equivalents. Currently, LLM outputs often need to be edited and evaluated by a human to ensure both factual accuracy and comprehensibility, and this is true for the above use case. We aim to minimize this step by proposing an agentic workflow with the Reflexion framework, which uses iterative self-reflection to correct outputs from an LLM. This pipeline was tested and compared to zero-shot prompting on 16 randomized radiology reports. In our multi-agent approach, reports had an accuracy rate of 94.94% when looking at verification of ICD-10 codes, compared to zero-shot prompted reports, which had an accuracy rate of 68.23%. Additionally, 81.25% of the final reflected reports required no corrections for accuracy or readability, while only 25% of zero-shot prompted reports met these criteria without needing modifications. These results indicate that our approach presents a feasible method for communicating clinical findings to patients in a quick, efficient and coherent manner whilst also retaining medical accuracy. The codebase is available for viewing at http://github.com/malavikhasudarshan/Multi-Agent-Patient-Letter-Generation., Comment: 12 pages, 7 figures

Determining the best attainable threshold for qudit magic state distillation is directly related to the question of whether or not contextuality is sufficient for universal quantum computation. We carry out a search for high-threshold magic state distillation routines for a highly-symmetric qutrit magic state known as the strange state. Our search covers a large class of $[[n,1]]_3$ qutrit stabilizer codes with up to 23 qutrits, and is facilitated by a theorem that relates the distillation performance of a qudit stabilizer code to its weight-enumerators. We could not find any code with $n<23$ qutrits that distills the strange state with better than linear noise suppression, other than the 11-qutrit Golay code. However, for $n=23$, we find over 600 CSS codes that can distill the qutrit strange state with cubic noise suppression. While none of these codes surpass the threshold of the 11-qutrit Golay code, their existence suggests that, for large codes, the ability to distill the qutrit strange state is somewhat generic., Comment: 27 pages, 5 figures, one ancillary file

Modern artificial intelligence (AI) systems are powered by foundation models. This paper presents a new set of foundation models, called Llama 3. It is a herd of language models that natively support multilinguality, coding, reasoning, and tool usage. Our largest model is a dense Transformer with 405B parameters and a context window of up to 128K tokens. This paper presents an extensive empirical evaluation of Llama 3. We find that Llama 3 delivers comparable quality to leading language models such as GPT-4 on a plethora of tasks. We publicly release Llama 3, including pre-trained and post-trained versions of the 405B parameter language model and our Llama Guard 3 model for input and output safety. The paper also presents the results of experiments in which we integrate image, video, and speech capabilities into Llama 3 via a compositional approach. We observe this approach performs competitively with the state-of-the-art on image, video, and speech recognition tasks. The resulting models are not yet being broadly released as they are still under development.

Large Language Models (LLMs) have significantly transformed the educational landscape, offering new tools for students, instructors, and teaching assistants. This paper investigates the application of LLMs in assisting teaching assistants (TAs) with viva and code assessments in an advanced computing class on distributed systems in an Indian University. We develop TAMIGO, an LLM-based system for TAs to evaluate programming assignments. For viva assessment, the TAs generated questions using TAMIGO and circulated these questions to the students for answering. The TAs then used TAMIGO to generate feedback on student answers. For code assessment, the TAs selected specific code blocks from student code submissions and fed it to TAMIGO to generate feedback for these code blocks. The TAMIGO-generated feedback for student answers and code blocks was used by the TAs for further evaluation. We evaluate the quality of LLM-generated viva questions, model answers, feedback on viva answers, and feedback on student code submissions. Our results indicate that LLMs are highly effective at generating viva questions when provided with sufficient context and background information. However, the results for LLM-generated feedback on viva answers were mixed; instances of hallucination occasionally reduced the accuracy of feedback. Despite this, the feedback was consistent, constructive, comprehensive, balanced, and did not overwhelm the TAs. Similarly, for code submissions, the LLM-generated feedback was constructive, comprehensive and balanced, though there was room for improvement in aligning the feedback with the instructor-provided rubric for code evaluation. Our findings contribute to understanding the benefits and limitations of integrating LLMs into educational settings., Comment: Under review

Reversing a diffusion process by learning its score forms the heart of diffusion-based generative modeling and for estimating properties of scientific systems. The diffusion processes that are tractable center on linear processes with a Gaussian stationary distribution. This limits the kinds of models that can be built to those that target a Gaussian prior or more generally limits the kinds of problems that can be generically solved to those that have conditionally linear score functions. In this work, we introduce a family of tractable denoising score matching objectives, called local-DSM, built using local increments of the diffusion process. We show how local-DSM melded with Taylor expansions enables automated training and score estimation with nonlinear diffusion processes. To demonstrate these ideas, we use automated-DSM to train generative models using non-Gaussian priors on challenging low dimensional distributions and the CIFAR10 image dataset. Additionally, we use the automated-DSM to learn the scores for nonlinear processes studied in statistical physics.

In order to communicate a message over a noisy channel, a sender (Alice) uses an error-correcting code to encode her message $x$ into a codeword. The receiver (Bob) decodes it correctly whenever there is at most a small constant fraction of adversarial error in the transmitted codeword. This work investigates the setting where Bob is computationally bounded. Specifically, Bob receives the message as a stream and must process it and write $x$ in order to a write-only tape while using low (say polylogarithmic) space. We show three basic results about this setting, which are informally as follows: (1) There is a stream decodable code of near-quadratic length. (2) There is no stream decodable code of sub-quadratic length. (3) If Bob need only compute a private linear function of the input bits, instead of writing them all to the output tape, there is a stream decodable code of near-linear length.

Recurrent large language models that compete with Transformers in language modeling perplexity are emerging at a rapid rate (e.g., Mamba, RWKV). Excitingly, these architectures use a constant amount of memory during inference. However, due to the limited memory, recurrent LMs cannot recall and use all the information in long contexts leading to brittle in-context learning (ICL) quality. A key challenge for efficient LMs is selecting what information to store versus discard. In this work, we observe the order in which information is shown to the LM impacts the selection difficulty. To formalize this, we show that the hardness of information recall reduces to the hardness of a problem called set disjointness (SD), a quintessential problem in communication complexity that requires a streaming algorithm (e.g., recurrent model) to decide whether inputted sets are disjoint. We empirically and theoretically show that the recurrent memory required to solve SD changes with set order, i.e., whether the smaller set appears first in-context. Our analysis suggests, to mitigate the reliance on data order, we can put information in the right order in-context or process prompts non-causally. Towards that end, we propose: (1) JRT-Prompt, where context gets repeated multiple times in the prompt, effectively showing the model all data orders. This gives $11.0 \pm 1.3$ points of improvement, averaged across $16$ recurrent LMs and the $6$ ICL tasks, with $11.9\times$ higher throughput than FlashAttention-2 for generation prefill (length $32$k, batch size $16$, NVidia H100). We then propose (2) JRT-RNN, which uses non-causal prefix-linear-attention to process prompts and provides $99\%$ of Transformer quality at $360$M params., $30$B tokens and $96\%$ at $1.3$B params., $50$B tokens on average across the tasks, with $19.2\times$ higher throughput for prefill than FA2.

Measuring a qubit state is a fundamental yet error-prone operation in quantum computing. These errors can arise from various sources, such as crosstalk, spontaneous state transitions, and excitations caused by the readout pulse. Here, we utilize an integrated approach to deploy neural networks onto field-programmable gate arrays (FPGA). We demonstrate that implementing a fully connected neural network accelerator for multi-qubit readout is advantageous, balancing computational complexity with low latency requirements without significant loss in accuracy. The neural network is implemented by quantizing weights, activation functions, and inputs. The hardware accelerator performs frequency-multiplexed readout of five superconducting qubits in less than 50 ns on a radio frequency system on chip (RFSoC) ZCU111 FPGA, marking the advent of RFSoC-based low-latency multi-qubit readout using neural networks. These modules can be implemented and integrated into existing quantum control and readout platforms, making the RFSoC ZCU111 ready for experimental deployment., Comment: 10 pages, 6 figures

The spherical mean transform associates to a function $f$ its integral averages over all spheres. We consider the spherical mean transform for functions supported in the unit ball $\mathbb{B}$ in $\mathbb{R}^n$ for odd $n$, with the centers of integration spheres restricted to the unit sphere $\mathbb{S}^{n-1}$. In this setup, Rubin employed properties of Erd\'elyi-Kober fractional integrals and analytic continuation to re-derive the explicit inversion formulas proved earlier by Finch, Patch, and Rakesh using wave equation techniques. As part of his work, Rubin stated a conjecture relating spherical mean transform, its associated backprojection operator and the Riesz potential. Furthermore, he pointed to the necessity of a detailed analysis of injectivity of the backprojection operator as a crucial step toward the resolution of his conjecture. This article addresses both questions posed by Rubin by providing a characterization of the null space of the backprojection operator, and disproving the conjecture through the construction of an explicit counterexample. Crucial to the proofs is the range characterization for the spherical mean transform in odd dimensions derived recently by the authors.

We present the results of a nation-wide baseline survey, conducted by us, for the status of Astronomy education among secondary school students in India. The survey was administered in 10 different languages to over 2000 students from diverse backgrounds, and it explored multiple facets of their perspectives on astronomy. The topics included students' views on the incorporation of astronomy in curricula, their grasp of fundamental astronomical concepts, access to educational resources, cultural connections to astronomy, and their levels of interest and aspirations in the subject. We find notable deficiencies in students' knowledge of basic astronomical principles, with only a minority demonstrating proficiency in key areas such as celestial sizes, distances, and lunar phases. Furthermore, access to resources such as telescopes and planetariums remain limited across the country. Despite these challenges, a significant majority of students expressed a keen interest in astronomy. We further analyze the data along socioeconomic and gender lines. Particularly striking were the socioeconomic disparities, with students from resource-poor backgrounds often having lower levels of access and proficiency. Some differences were observed between genders, although not very pronounced. The insights gleaned from this study hold valuable implications for the development of a more robust astronomy curriculum and the design of effective teacher training programs in the future., Comment: 15 pages, 19 figures

The observation of several L-band emission sources in the S cluster has led to a rich discussion of their nature. However, a definitive answer to the classification of the dusty objects requires an explanation for the detection of compact Doppler-shifted Br$\gamma$ emission. The ionized hydrogen in combination with the observation of mid-infrared $L$-band continuum emission suggests that most of these sources are embedded in a dusty envelope. These embedded sources are part of the S-cluster, and their relationship to the S-stars is still under debate. Until now, the question of the origin of these two populations is vague, although all explanations favor migration processes for the individual cluster members. This work revisits the S-cluster and its dusty members orbiting the supermassive black hole Sgr~A* on bound Keplerian orbits from a kinematic perspective. The aim is to explore the Keplerian parameters for patterns that might imply a non-random distribution of the sample. Additionally, various analytical aspects are considered to address the nature of the dusty sources. Based on the photometric analysis, we estimated the individual $H-K$ and $K-L$ colors for the source sample and compared the results to known cluster members. The classification revealed a noticeable contrast between the S-stars and the dusty sources. To fit the flux-density distribution, we utilized the radiative transfer code HYPERION and implemented a Young Stellar Object Class I model. We obtained the position angle from the Keplerian fit results, and additionally, we analyzed the distribution of the inclinations and the longitudes of the ascending node. The colors of the dusty sources suggest a stellar nature consistent with the spectral energy distribution in the near and mid-infrared domains., Comment: 36 pages, 32 figures, accepted and published by A&A

Given a family of cyclic covers of $\mathbb{P}^1$ and a prime $p$ of good reduction, by [12] the generic Newton polygon (resp. Ekedahl--Oort type) in the family ($\mu$-ordinary) is known. In this paper, we investigate the existence of non-$\mu$-ordinary smooth curves in the family. In particular, under some auxiliary conditions, we show that when $p$ is sufficiently large the complement of the $\mu$-ordinary locus is always non empty, and for $1$-dimensional families with condition on signature type, we obtain a lower bound for the number of non-$\mu$-ordinary smooth curves. In specific examples, for small $m$, the above general statement can be improved, and we establish the non emptiness of all codimension 1 non-$\mu$-ordinary Newton/Ekedahl--Oort strata ({\em almost} $\mu$-ordinary). Our method relies on further study of the extended Hasse-Witt matrix initiated in [12].

Magnetic Resonance (MR) imaging, despite its proven diagnostic utility, remains an inaccessible imaging modality for disease surveillance at the population level. A major factor rendering MR inaccessible is lengthy scan times. An MR scanner collects measurements associated with the underlying anatomy in the Fourier space, also known as the k-space. Creating a high-fidelity image requires collecting large quantities of such measurements, increasing the scan time. Traditionally to accelerate an MR scan, image reconstruction from under-sampled k-space data is the method of choice. However, recent works show the feasibility of bypassing image reconstruction and directly learning to detect disease directly from a sparser learned subset of the k-space measurements. In this work, we propose Adaptive Sampling for MR (ASMR), a sampling method that learns an adaptive policy to sequentially select k-space samples to optimize for target disease detection. On 6 out of 8 pathology classification tasks spanning the Knee, Brain, and Prostate MR scans, ASMR reaches within 2% of the performance of a fully sampled classifier while using only 8% of the k-space, as well as outperforming prior state-of-the-art work in k-space sampling such as EMRT, LOUPE, and DPS., Comment: ICML 2024. Project website at https://adaptive-sampling-mr.github.io

Given a dataset of individuals each described by a covariate vector, a treatment, and an observed outcome on the treatment, the goal of the individual treatment effect (ITE) estimation task is to predict outcome changes resulting from a change in treatment. A fundamental challenge is that in the observational data, a covariate's outcome is observed only under one treatment, whereas we need to infer the difference in outcomes under two different treatments. Several existing approaches address this issue through training with inferred pseudo-outcomes, but their success relies on the quality of these pseudo-outcomes. We propose PairNet, a novel ITE estimation training strategy that minimizes losses over pairs of examples based on their factual observed outcomes. Theoretical analysis for binary treatments reveals that PairNet is a consistent estimator of ITE risk, and achieves smaller generalization error than baseline models. Empirical comparison with thirteen existing methods across eight benchmarks, covering both discrete and continuous treatments, shows that PairNet achieves significantly lower ITE error compared to the baselines. Also, it is model-agnostic and easy to implement., Comment: Lokesh and Pranava contributed equally. Accepted at ICML-24

The multi-valued byzantine agreement protocol (MVBA) in the authenticated setting has been widely used as a core to design atomic broadcast and fault-tolerant state machine replication protocols in asynchronous networks. Originating from the seminal work of Cachin et al. \cite{CACHIN01}, subsequent research endeavors have sought to optimize protocol efficiency in terms of communication complexity. Notable advancements following Cachin's contributions include: i) VABA \cite{BYZ17}, requiring multiple protocol instances to achieve agreement on a party's request, and ii) Dumbo-MVBA \cite{LU20}, employing a cryptographic asynchronous dispersal and recovery methods to manage communication complexity alongside additional computational and communication rounds overheads. Our objective is to devise an MVBA protocol that achieves agreement in each instance without extra computation and communication rounds while maintaining the optimal metrics. Central to our design approach is the introduction of the committee in the classic MVBA protocol, wherein a randomly selected subset of ($f+1$, where $n=3f+1$) parties get selected and simultaneously broadcast their requests (transactions) to gather verifiable proofs. Successive distributions of these proofs afford us the necessary properties to employ the asynchronous binary Byzantine agreement (ABBA) protocol for reaching an agreement on a selected party's requests. By integrating the committee and ABBA protocols, we devise the optimal MVBA protocol, termed pMVBA (Prioritized-MVBA). This protocol exhibits resilience to tolerate up to $\lfloor \frac{n}{3}\rfloor$ Byzantine failures, with an expected runtime of $O(1)$, optimal message complexity of $O(n^2)$, and optimal communication complexity $O((l+\lambda)n^2)$ ., Comment: 23 pages

Deep Neural Networks (DNN) have become a promising paradigm when developing Artificial Intelligence (AI) and Machine Learning (ML) applications. However, DNN applications are vulnerable to fake data that are crafted with adversarial attack algorithms. Under adversarial attacks, the prediction accuracy of DNN applications suffers, making them unreliable. In order to defend against adversarial attacks, we introduce a novel noise-reduction procedure, Vector Quantization U-Net (VQUNet), to reduce adversarial noise and reconstruct data with high fidelity. VQUNet features a discrete latent representation learning through a multi-scale hierarchical structure for both noise reduction and data reconstruction. The empirical experiments show that the proposed VQUNet provides better robustness to the target DNN models, and it outperforms other state-of-the-art noise-reduction-based defense methods under various adversarial attacks for both Fashion-MNIST and CIFAR10 datasets. When there is no adversarial attack, the defense method has less than 1% accuracy degradation for both datasets., Comment: 8 pages, 6 figures

Nonlinearities can have a profound influence on the dynamics and equilibrium properties of discrete lattice systems. The simple case of two coupled modes with self-nonlinearities gives rise to the rich bosonic Josephson effects. In many-site arrays, nonlinearities yield a wealth of rich phenomena, including a variety of solitonic excitations, the emergence of vortex lattices in the presence of gauge fields, and the general support of chaotic dynamics. Here, we experimentally explore a three-site mechanical ring with tunable gauge fields and nonlinearities. We observe a macroscopic self-trapping transition that is tunable by the magnetic flux, consistent with the equilibrium response. We further observe novel behavior that appears only out of equilibrium, the emergence of interaction-stabilized chiral solitary waves. These results provide a starting point to explore nonlinear phenomena arising in larger mechanical arrays coupled to static and dynamical gauge fields., Comment: 6 pages, 3 figures