Recognizing if LLM output can be grounded in evidence is central to many tasks in NLP: retrieval-augmented generation, summarization, document-grounded dialogue, and more. Current approaches to this kind of "fact-checking" are based on verifying each piece of a model generation against potential evidence using an LLM. However, this process can be very computationally expensive, requiring many calls to LLMs to check a single response. In this work, we show how to build small models that have GPT-4-level performance but for 400x lower cost. We do this by constructing synthetic training data with GPT-4, which involves creating realistic yet challenging instances of factual errors via a structured generation procedure. Training on this data teaches models to check each fact in the claim and recognize synthesis of information across sentences. For evaluation, we unify pre-existing datasets into a benchmark LLM-AggreFact, collected from recent work on fact-checking and grounding LLM generations. Our best system MiniCheck-FT5 (770M parameters) outperforms all systems of comparable size and reaches GPT-4 accuracy. We release LLM-AggreFact, code for data synthesis, and models.

Motivated by applications of large embedding models, we study differentially private (DP) optimization problems under sparsity of individual gradients. We start with new near-optimal bounds for the classic mean estimation problem but with sparse data, improving upon existing algorithms particularly for the high-dimensional regime. Building on this, we obtain pure- and approximate-DP algorithms with almost optimal rates for stochastic convex optimization with sparse gradients; the former represents the first nearly dimension-independent rates for this problem. Finally, we study the approximation of stationary points for the empirical loss in approximate-DP optimization and obtain rates that depend on sparsity instead of dimension, modulo polylogarithmic factors.

Abstract--Through the implementation of reconfigurability to achieve flexibility and adaptation to tasks by morphology changes rather than by increasing the number of joints, malleable robots present advantages over traditional serial robot arms in regards to reduced weight, size, and cost. While limited in degrees of freedom (DOF), malleable robots still provide versatility across operations typically served by systems using higher DOF than required by the tasks. In this paper, we present the creation of a 2-DOF malleable robot, detailing the design of joints and malleable link, along with its modelling through forward and inverse kinematics, and a reconfiguration methodology that informs morphology changes based on end effector location-- determining how the user should reshape the robot to enable a task previously unattainable. The recalibration and motion planning for making robot motion possible after reconfiguration are also discussed, and thorough experiments with the prototype to evaluate accuracy and reliability of the system are presented. ECONFIGURABLE robot systems provide several key potential advantages over traditional robots, including of the robot (such as locomotion), albeit with a decrease in increased task versatility by adapting to better suit tasks, the performance for a specific task compared to a specialised and reduced robot cost due to a smaller total number of robot. While the majority of reconfigurable robots are modular, modules, such as links and joints. As such, there has been reconfiguration can also be achieved by locking aspects of significant research into the development of reconfigurable the robot. Examples include directly locking revolute joints to robots, with the most popular approach utilising modularity reduce the DOF of the robot [11], and locking passive cylindrical as the method of reconfiguration, as this allows for the joints carefully positioned to directly vary the Denavit-interchangeability of parts, leading to self-repair [1], [2].

In this study, we introduce SpamDam, a SMS spam detection framework designed to overcome key challenges in detecting and understanding SMS spam, such as the lack of public SMS spam datasets, increasing privacy concerns of collecting SMS data, and the need for adversary-resistant detection models. SpamDam comprises four innovative modules: an SMS spam radar that identifies spam messages from online social networks(OSNs); an SMS spam inspector for statistical analysis; SMS spam detectors(SSDs) that enable both central training and federated learning; and an SSD analyzer that evaluates model resistance against adversaries in realistic scenarios. Leveraging SpamDam, we have compiled over 76K SMS spam messages from Twitter and Weibo between 2018 and 2023, forming the largest dataset of its kind. This dataset has enabled new insights into recent spam campaigns and the training of high-performing binary and multi-label classifiers for spam detection. Furthermore, effectiveness of federated learning has been well demonstrated to enable privacy-preserving SMS spam detection. Additionally, we have rigorously tested the adversarial robustness of SMS spam detection models, introducing the novel reverse backdoor attack, which has shown effectiveness and stealthiness in practical tests.

State-of-the-art large language models (LLMs) have become indispensable tools for various tasks. However, training LLMs to serve as effective assistants for humans requires careful consideration. A promising approach is reinforcement learning from human feedback (RLHF), which leverages human feedback to update the model in accordance with human preferences and mitigate issues like toxicity and hallucinations. Yet, an understanding of RLHF for LLMs is largely entangled with initial design choices that popularized the method and current research focuses on augmenting those choices rather than fundamentally improving the framework. In this paper, we analyze RLHF through the lens of reinforcement learning principles to develop an understanding of its fundamentals, dedicating substantial focus to the core component of RLHF -- the reward model. Our study investigates modeling choices, caveats of function approximation, and their implications on RLHF training algorithms, highlighting the underlying assumptions made about the expressivity of reward. Our analysis improves the understanding of the role of reward models and methods for their training, concurrently revealing limitations of the current methodology. We characterize these limitations, including incorrect generalization, model misspecification, and the sparsity of feedback, along with their impact on the performance of a language model. The discussion and analysis are substantiated by a categorical review of current literature, serving as a reference for researchers and practitioners to understand the challenges of RLHF and build upon existing efforts.

We introduce two distinct types of network pruning within our study: 1) global pruning, which extends from server to client, and 2) local pruning, where each client's network is pruned based on its own specific data. In our setting, we assume federated pruning is the scenario with both possible global and local pruning. Federated network pruning, a closely related field, pursues the objective of identifying the optimal or near-optimal pruned neural network at each communication from the server to the clients, as documented in works of Jiang et al. (2022a) and Huang et al. (2022), for example. During the initial phase of global pruning, (Jiang et al., 2022a) isolates a single potent and reliable client to initiate model pruning. The subsequent stage of local pruning incorporates all clients, advancing the adaptive pruning process. This process involves not only parameter removal but also the reintroduction of parameters, complemented by the standard FedAvg (McMahan et al., 2017). However, the need for substantial local memory to record the updated relevance measures of all parameters in the full-scale model poses a challenge. As a solution to this problem, Huang et al. (2022) proposes an adaptive batch normalization and progressive pruning modules that utilize sparse local computation. Yet, these methods overlook explicit considerations for constraints related to client-side computational resources and communication bandwidth.

There is a belief that learning to compress well will lead to intelligence. Recently, language modeling has been shown to be equivalent to compression, which offers a compelling rationale for the success of large language models (LLMs): the development of more advanced language models is essentially enhancing compression which facilitates intelligence. Despite such appealing discussions, little empirical evidence is present for the interplay between compression and intelligence. In this work, we examine their relationship in the context of LLMs, treating LLMs as data compressors. Given the abstract concept of "intelligence", we adopt the average downstream benchmark scores as a surrogate, specifically targeting intelligence related to knowledge and commonsense, coding, and mathematical reasoning. Across 12 benchmarks, our study brings together 30 public LLMs that originate from diverse organizations. Remarkably, we find that LLMs' intelligence -- reflected by average benchmark scores -- almost linearly correlates with their ability to compress external text corpora. These results provide concrete evidence supporting the belief that superior compression indicates greater intelligence. Furthermore, our findings suggest that compression efficiency, as an unsupervised metric derived from raw text corpora, serves as a reliable evaluation measure that is linearly associated with the model capabilities. We open-source our compression datasets as well as our data collection pipelines to facilitate future researchers to assess compression properly.

Software vulnerabilities are a fundamental reason for the prevalence of cyber attacks and their identification is a crucial yet challenging problem in cyber security. In this paper, we apply and compare different machine learning algorithms for source code vulnerability detection specifically for Python programming language. Our experimental evaluation demonstrates that our Bidirectional Long Short-Term Memory (BiLSTM) model achieves a remarkable performance (average Accuracy = 98.6%, average F-Score = 94.7%, average Precision = 96.2%, average Recall = 93.3%, average ROC = 99.3%), thereby, establishing a new benchmark for vulnerability detection in Python source code.

The rapid developments of various machine learning models and their deployments in several applications has led to discussions around the importance of looking beyond the accuracies of these models. Fairness of such models is one such aspect that is deservedly gaining more attention. In this work, we analyse the natural language representations of documents and sentences (i.e., encodings) for any embedding-level bias that could potentially also affect the fairness of the downstream tasks that rely on them. We identify bias in these encodings either towards or against different sub-groups based on the difference in their reconstruction errors along various subsets of principal components. We explore and recommend ways to mitigate such bias in the encodings while also maintaining a decent accuracy in classification models that use them.

The emergence of Large Multimodal Models (LMMs) marks a significant milestone in the development of artificial intelligence. Insurance, as a vast and complex discipline, involves a wide variety of data forms in its operational processes, including text, images, and videos, thereby giving rise to diverse multimodal tasks. Despite this, there has been limited systematic exploration of multimodal tasks specific to insurance, nor a thorough investigation into how LMMs can address these challenges. In this paper, we explore GPT-4V's capabilities in the insurance domain. We categorize multimodal tasks by focusing primarily on visual aspects based on types of insurance (e.g., auto, household/commercial property, health, and agricultural insurance) and insurance stages (e.g., risk assessment, risk monitoring, and claims processing). Our experiment reveals that GPT-4V exhibits remarkable abilities in insurance-related tasks, demonstrating not only a robust understanding of multimodal content in the insurance domain but also a comprehensive knowledge of insurance scenarios. However, there are notable shortcomings: GPT-4V struggles with detailed risk rating and loss assessment, suffers from hallucination in image understanding, and shows variable support for different languages. Through this work, we aim to bridge the insurance domain with cutting-edge LMM technology, facilitate interdisciplinary exchange and development, and provide a foundation for the continued advancement and evolution of future research endeavors.