Large Language Models (LLMs) have shown remarkable proficiency in following instructions, making them valuable in customer-facing applications. However, their impressive capabilities also raise concerns about the amplification of risks posed by adversarial instructions, which can be injected into the model input by third-party attackers to manipulate LLMs' original instructions and prompt unintended actions and content. Therefore, it is crucial to understand LLMs' ability to accurately discern which instructions to follow to ensure their safe deployment in real-world scenarios. In this paper, we propose a pioneering benchmark for automatically evaluating the robustness of instruction-following LLMs against adversarial instructions injected in the prompt. The objective of this benchmark is to quantify the extent to which LLMs are influenced by injected adversarial instructions and assess their ability to differentiate between these injected adversarial instructions and original user instructions. Through experiments conducted with state-of-the-art instruction-following LLMs, we uncover significant limitations in their robustness against adversarial instruction injection attacks. Furthermore, our findings indicate that prevalent instruction-tuned models are prone to being ``overfitted'' to follow any instruction phrase in the prompt without truly understanding which instructions should be followed. This highlights the need to address the challenge of training models to comprehend prompts instead of merely following instruction phrases and completing the text. The data and code can be found at \url{//github.com/Leezekun/Adv-Instruct-Eval}.
相關內容
Large Language Models (LLMs) employing Chain-of-Thought (CoT) prompting have broadened the scope for improving multi-step reasoning capabilities. Usually, answer calibration strategies such as step-level or path-level calibration play a vital role in multi-step reasoning. While effective, there remains a significant gap in our understanding of the key factors that drive their success. In this paper, we break down the design of recent answer calibration strategies and present a unified view which establishes connections between them. We then conduct a thorough evaluation on these strategies from a unified view, systematically scrutinizing step-level and path-level answer calibration across multiple paths. Our study holds the potential to illuminate key insights for optimizing multi-step reasoning with answer calibration.
Deep Neural Networks (DNNs) are extremely computationally demanding, which presents a large barrier to their deployment on resource-constrained devices. Since such devices are where many emerging deep learning applications lie (e.g., drones, vision-based medical technology), significant bodies of work from both the machine learning and systems communities have attempted to provide optimizations to accelerate DNNs. To help unify these two perspectives, in this paper we combine machine learning and systems techniques within the Deep Learning Acceleration Stack (DLAS), and demonstrate how these layers can be tightly dependent on each other with an across-stack perturbation study. We evaluate the impact on accuracy and inference time when varying different parameters of DLAS across two datasets, seven popular DNN architectures, four DNN compression techniques, three algorithmic primitives with sparse and dense variants, untuned and auto-scheduled code generation, and four hardware platforms. Our evaluation highlights how perturbations across DLAS parameters can cause significant variation and across-stack interactions. The highest level observation from our evaluation is that the model size, accuracy, and inference time are not guaranteed to be correlated. Overall we make 13 key observations, including that speedups provided by compression techniques are very hardware dependent, and that compiler auto-tuning can significantly alter what the best algorithm to use for a given configuration is. With DLAS, we aim to provide a reference framework to aid machine learning and systems practitioners in reasoning about the context in which their respective DNN acceleration solutions exist in. With our evaluation strongly motivating the need for co-design, we believe that DLAS can be a valuable concept for exploring the next generation of co-designed accelerated deep learning solutions.
Machine Translation (MT) continues to improve in quality and adoption, yet the inadvertent perpetuation of gender bias remains a significant concern. Despite numerous studies into gender bias in translations from gender-neutral languages such as Turkish into more strongly gendered languages like English, there are no benchmarks for evaluating this phenomenon or for assessing mitigation strategies. To address this gap, we introduce GATE X-E, an extension to the GATE (Rarrick et al., 2023) corpus, that consists of human translations from Turkish, Hungarian, Finnish, and Persian into English. Each translation is accompanied by feminine, masculine, and neutral variants for each possible gender interpretation. The dataset, which contains between 1250 and 1850 instances for each of the four language pairs, features natural sentences with a wide range of sentence lengths and domains, challenging translation rewriters on various linguistic phenomena. Additionally, we present an English gender rewriting solution built on GPT-3.5 Turbo and use GATE X-E to evaluate it. We open source our contributions to encourage further research on gender debiasing.
Language models (LMs) have demonstrated remarkable capabilities across a wide range of tasks in various domains. Despite their impressive performance, the reliability of their output is concerning and questionable regarding the demand for AI safety. Assessing the confidence of LM predictions and calibrating them across different tasks with the aim to align LM confidence with accuracy can help mitigate risks and enable LMs to make better decisions. There have been various works in this respect, but there has been no comprehensive overview of this important research area. The present survey aims to bridge this gap. In particular, we discuss methods and techniques for LM confidence estimation and calibration, encompassing different LMs and various tasks. We further outline the challenges of estimating the confidence for large language models and we suggest some promising directions for future work.
The remarkable ability of Large Language Models (LLMs) to understand and follow instructions has sometimes been limited by their in-context learning (ICL) performance in low-resource languages. To address this, we introduce a novel approach that leverages cross-lingual retrieval-augmented in-context learning (CREA-ICL). By extracting semantically similar prompts from high-resource languages, we aim to improve the zero-shot performance of multilingual pre-trained language models (MPLMs) across diverse tasks. Though our approach yields steady improvements in classification tasks, it faces challenges in generation tasks. Our evaluation offers insights into the performance dynamics of retrieval-augmented in-context learning across both classification and generation domains.
Large Language Models (LLMs) can solve complex reasoning tasks by generating rationales for their predictions. Distilling these capabilities into a smaller, compact model can facilitate the creation of specialized, cost-effective models tailored for specific tasks. However, smaller models often face challenges in complex reasoning tasks and often deviate from the correct reasoning path. We show that LLMs can guide smaller models and bring them back to the correct reasoning path only if they intervene at the right time. We show that smaller models fail to reason primarily due to their difficulty in initiating the process, and that guiding them in the right direction can lead to a performance gain of over 100%. We explore different model sizes and evaluate the benefits of providing guidance to improve reasoning in smaller models.
The increasing success of Large Language Models (LLMs) in variety of tasks lead to their widespread use in our lives which necessitates the examination of these models from different perspectives. The alignment of these models to human values is an essential concern in order to establish trust that we have safe and responsible systems. In this paper, we aim to find out which values and principles are embedded in LLMs in the process of moral justification. For this purpose, we come up with three different moral perspective categories: Western tradition perspective (WT), Abrahamic tradition perspective (AT), and Spiritualist/Mystic tradition perspective (SMT). In two different experiment settings, we asked models to choose principles from the three for suggesting a moral action and evaluating the moral permissibility of an action if one tries to justify an action on these categories, respectively. Our experiments indicate that tested LLMs favors the Western tradition moral perspective over others. Additionally, we observe that there potentially exists an over-alignment towards religious values represented in the Abrahamic Tradition, which causes models to fail to recognize an action is immoral if it is presented as a "religious-action". We believe that these results are essential in order to direct our attention in future efforts.
Large Language Models (LLMs) have shown remarkable proficiency in language understanding and have been successfully applied to a variety of real-world tasks through task-specific fine-tuning or prompt engineering. Despite these advancements, it remains an open question whether LLMs are fundamentally capable of reasoning and planning, or if they primarily rely on recalling and synthesizing information from their training data. In our research, we introduce a novel task -- Minesweeper -- specifically designed in a format unfamiliar to LLMs and absent from their training datasets. This task challenges LLMs to identify the locations of mines based on numerical clues provided by adjacent opened cells. Successfully completing this task requires an understanding of each cell's state, discerning spatial relationships between the clues and mines, and strategizing actions based on logical deductions drawn from the arrangement of the cells. Our experiments, including trials with the advanced GPT-4 model, indicate that while LLMs possess the foundational abilities required for this task, they struggle to integrate these into a coherent, multi-step logical reasoning process needed to solve Minesweeper. These findings highlight the need for further research to understand and nature of reasoning capabilities in LLMs under similar circumstances, and to explore pathways towards more sophisticated AI reasoning and planning models.
Large Language Models (LLMs) have achieved remarkable success in many formal language oriented tasks, such as structural data-to-text and semantic parsing. However current benchmarks mostly follow the data distribution of the pre-training data of LLMs. Therefore, a natural question rises that do LLMs really understand the structured semantics of formal languages. In this paper, we investigate this problem on a special case, converse binary relation. We introduce a new benchmark ConvRe focusing on converse relations, which contains 17 relations and 1240 triples extracted from popular knowledge graph completion datasets. Our ConvRE features two tasks, Re2Text and Text2Re, which are formulated as multi-choice question answering to evaluate LLMs' ability to determine the matching between relations and associated text. For the evaluation protocol, apart from different prompting methods, we further introduce variants to the test text and few-shot example text. We conduct experiments on three popular LLM families and have observed various scaling trends. The results suggest that LLMs often resort to shortcut learning and still face challenges on our proposed benchmark.
In pace with developments in the research field of artificial intelligence, knowledge graphs (KGs) have attracted a surge of interest from both academia and industry. As a representation of semantic relations between entities, KGs have proven to be particularly relevant for natural language processing (NLP), experiencing a rapid spread and wide adoption within recent years. Given the increasing amount of research work in this area, several KG-related approaches have been surveyed in the NLP research community. However, a comprehensive study that categorizes established topics and reviews the maturity of individual research streams remains absent to this day. Contributing to closing this gap, we systematically analyzed 507 papers from the literature on KGs in NLP. Our survey encompasses a multifaceted review of tasks, research types, and contributions. As a result, we present a structured overview of the research landscape, provide a taxonomy of tasks, summarize our findings, and highlight directions for future work.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191