Pub Date : 2024-02-15DOI: 10.48550/arXiv.2402.09880
Timothy R. McIntosh, Teo Susnjak, Tong Liu, Paul Watters, M. Halgamuge
The rapid rise in popularity of Large Language Models (LLMs) with emerging capabilities has spurred public curiosity to evaluate and compare different LLMs, leading many researchers to propose their LLM benchmarks. Noticing preliminary inadequacies in those benchmarks, we embarked on a study to critically assess 23 state-of-the-art LLM benchmarks, using our novel unified evaluation framework through the lenses of people, process, and technology, under the pillars of functionality and security. Our research uncovered significant limitations, including biases, difficulties in measuring genuine reasoning, adaptability, implementation inconsistencies, prompt engineering complexity, evaluator diversity, and the overlooking of cultural and ideological norms in one comprehensive assessment. Our discussions emphasized the urgent need for standardized methodologies, regulatory certainties, and ethical guidelines in light of Artificial Intelligence (AI) advancements, including advocating for an evolution from static benchmarks to dynamic behavioral profiling to accurately capture LLMs' complex behaviors and potential risks. Our study highlighted the necessity for a paradigm shift in LLM evaluation methodologies, underlining the importance of collaborative efforts for the development of universally accepted benchmarks and the enhancement of AI systems' integration into society.
{"title":"Inadequacies of Large Language Model Benchmarks in the Era of Generative Artificial Intelligence","authors":"Timothy R. McIntosh, Teo Susnjak, Tong Liu, Paul Watters, M. Halgamuge","doi":"10.48550/arXiv.2402.09880","DOIUrl":"https://doi.org/10.48550/arXiv.2402.09880","url":null,"abstract":"The rapid rise in popularity of Large Language Models (LLMs) with emerging capabilities has spurred public curiosity to evaluate and compare different LLMs, leading many researchers to propose their LLM benchmarks. Noticing preliminary inadequacies in those benchmarks, we embarked on a study to critically assess 23 state-of-the-art LLM benchmarks, using our novel unified evaluation framework through the lenses of people, process, and technology, under the pillars of functionality and security. Our research uncovered significant limitations, including biases, difficulties in measuring genuine reasoning, adaptability, implementation inconsistencies, prompt engineering complexity, evaluator diversity, and the overlooking of cultural and ideological norms in one comprehensive assessment. Our discussions emphasized the urgent need for standardized methodologies, regulatory certainties, and ethical guidelines in light of Artificial Intelligence (AI) advancements, including advocating for an evolution from static benchmarks to dynamic behavioral profiling to accurately capture LLMs' complex behaviors and potential risks. Our study highlighted the necessity for a paradigm shift in LLM evaluation methodologies, underlining the importance of collaborative efforts for the development of universally accepted benchmarks and the enhancement of AI systems' integration into society.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"7 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139963773","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-15DOI: 10.48550/arXiv.2402.09963
Michael Hahn, Mark Rofin
Empirical studies have identified a range of learnability biases and limitations of transformers, such as a persistent difficulty in learning to compute simple formal languages such as PARITY, and a bias towards low-degree functions. However, theoretical understanding remains limited, with existing expressiveness theory either overpredicting or underpredicting realistic learning abilities. We prove that, under the transformer architecture, the loss landscape is constrained by the input-space sensitivity: Transformers whose output is sensitive to many parts of the input string inhabit isolated points in parameter space, leading to a low-sensitivity bias in generalization. We show theoretically and empirically that this theory unifies a broad array of empirical observations about the learning abilities and biases of transformers, such as their generalization bias towards low sensitivity and low degree, and difficulty in length generalization for PARITY. This shows that understanding transformers' inductive biases requires studying not just their in-principle expressivity, but also their loss landscape.
{"title":"Why are Sensitive Functions Hard for Transformers?","authors":"Michael Hahn, Mark Rofin","doi":"10.48550/arXiv.2402.09963","DOIUrl":"https://doi.org/10.48550/arXiv.2402.09963","url":null,"abstract":"Empirical studies have identified a range of learnability biases and limitations of transformers, such as a persistent difficulty in learning to compute simple formal languages such as PARITY, and a bias towards low-degree functions. However, theoretical understanding remains limited, with existing expressiveness theory either overpredicting or underpredicting realistic learning abilities. We prove that, under the transformer architecture, the loss landscape is constrained by the input-space sensitivity: Transformers whose output is sensitive to many parts of the input string inhabit isolated points in parameter space, leading to a low-sensitivity bias in generalization. We show theoretically and empirically that this theory unifies a broad array of empirical observations about the learning abilities and biases of transformers, such as their generalization bias towards low sensitivity and low degree, and difficulty in length generalization for PARITY. This shows that understanding transformers' inductive biases requires studying not just their in-principle expressivity, but also their loss landscape.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"29 9","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139962134","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-15DOI: 10.48550/arXiv.2402.10175
Yinhong Liu, Yixuan Su, Ehsan Shareghi, Nigel Collier
Recent large language models (LLMs) have shown remarkable performance in aligning generated text with user intentions across various tasks. When it comes to long-form text generation, there has been a growing interest in generation from a discourse coherence perspective. However, existing lexical or semantic metrics such as BLEU, ROUGE, BertScore cannot effectively capture the discourse coherence. The development of discourse-specific automatic evaluation methods for assessing the output of LLMs warrants greater focus and exploration. In this paper, we present a novel automatic metric designed to quantify the discourse divergence between two long-form articles. Extensive experiments on three datasets from representative domains demonstrate that our metric aligns more closely with human preferences and GPT-4 coherence evaluation, outperforming existing evaluation methods.
{"title":"Unlocking Structure Measuring: Introducing PDD, an Automatic Metric for Positional Discourse Coherence","authors":"Yinhong Liu, Yixuan Su, Ehsan Shareghi, Nigel Collier","doi":"10.48550/arXiv.2402.10175","DOIUrl":"https://doi.org/10.48550/arXiv.2402.10175","url":null,"abstract":"Recent large language models (LLMs) have shown remarkable performance in aligning generated text with user intentions across various tasks. When it comes to long-form text generation, there has been a growing interest in generation from a discourse coherence perspective. However, existing lexical or semantic metrics such as BLEU, ROUGE, BertScore cannot effectively capture the discourse coherence. The development of discourse-specific automatic evaluation methods for assessing the output of LLMs warrants greater focus and exploration. In this paper, we present a novel automatic metric designed to quantify the discourse divergence between two long-form articles. Extensive experiments on three datasets from representative domains demonstrate that our metric aligns more closely with human preferences and GPT-4 coherence evaluation, outperforming existing evaluation methods.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"26 8","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139962160","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-15DOI: 10.48550/arXiv.2402.09757
Gobinda Ghosh, S. Majhi, Subhabrata Paul
The rapid progression in wireless communication technologies, especially in multicarrier code-division multiple access (MC-CDMA), there is a need of advanced code construction methods. Traditional approaches, mainly based on generalized Boolean functions, have limitations in code length versatility. This paper introduces a novel approach to constructing complete complementary codes (CCC) and Z-complementary code sets (ZCCS), for reducing interference in MC-CDMA systems. The proposed construction, distinct from Boolean function-based approaches, employs additive characters over Galois fields GF($p^{r}$), where $p$ is prime and $r$ is a positive integer. First, we develop CCCs with lengths of $p^{r}$, which are then extended to construct ZCCS with both unreported lengths and sizes of $np^{r}$, where $n$ are arbitrary positive integers. The versatility of this method is further highlighted as it includes the lengths of ZCCS reported in prior studies as special cases, underscoring the method's comprehensive nature and superiority.
{"title":"Construction of CCC and ZCCS Through Additive Characters Over Galois Field","authors":"Gobinda Ghosh, S. Majhi, Subhabrata Paul","doi":"10.48550/arXiv.2402.09757","DOIUrl":"https://doi.org/10.48550/arXiv.2402.09757","url":null,"abstract":"The rapid progression in wireless communication technologies, especially in multicarrier code-division multiple access (MC-CDMA), there is a need of advanced code construction methods. Traditional approaches, mainly based on generalized Boolean functions, have limitations in code length versatility. This paper introduces a novel approach to constructing complete complementary codes (CCC) and Z-complementary code sets (ZCCS), for reducing interference in MC-CDMA systems. The proposed construction, distinct from Boolean function-based approaches, employs additive characters over Galois fields GF($p^{r}$), where $p$ is prime and $r$ is a positive integer. First, we develop CCCs with lengths of $p^{r}$, which are then extended to construct ZCCS with both unreported lengths and sizes of $np^{r}$, where $n$ are arbitrary positive integers. The versatility of this method is further highlighted as it includes the lengths of ZCCS reported in prior studies as special cases, underscoring the method's comprehensive nature and superiority.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"20 7","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139962493","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-15DOI: 10.48550/arXiv.2402.09835
A. Feldmann, M. Lampis
In this paper we reassess the parameterized complexity and approximability of the well-studied Steiner Forest problem in several graph classes of bounded width. The problem takes an edge-weighted graph and pairs of vertices as input, and the aim is to find a minimum cost subgraph in which each given vertex pair lies in the same connected component. It is known that this problem is APX-hard in general, and NP-hard on graphs of treewidth 3, treedepth 4, and feedback vertex set size 2. However, Bateni, Hajiaghayi and Marx [JACM, 2011] gave an approximation scheme with a runtime of $n^{O(frac{k^2}{varepsilon})}$ on graphs of treewidth $k$. Our main result is a much faster efficient parameterized approximation scheme (EPAS) with a runtime of $2^{O(frac{k^2}{varepsilon} log frac{k^2}{varepsilon})} cdot n^{O(1)}$. If $k$ instead is the vertex cover number of the input graph, we show how to compute the optimum solution in $2^{O(k log k)} cdot n^{O(1)}$ time, and we also prove that this runtime dependence on $k$ is asymptotically best possible, under ETH. Furthermore, if $k$ is the size of a feedback edge set, then we obtain a faster $2^{O(k)} cdot n^{O(1)}$ time algorithm, which again cannot be improved under ETH.
{"title":"Parameterized Algorithms for Steiner Forest in Bounded Width Graphs","authors":"A. Feldmann, M. Lampis","doi":"10.48550/arXiv.2402.09835","DOIUrl":"https://doi.org/10.48550/arXiv.2402.09835","url":null,"abstract":"In this paper we reassess the parameterized complexity and approximability of the well-studied Steiner Forest problem in several graph classes of bounded width. The problem takes an edge-weighted graph and pairs of vertices as input, and the aim is to find a minimum cost subgraph in which each given vertex pair lies in the same connected component. It is known that this problem is APX-hard in general, and NP-hard on graphs of treewidth 3, treedepth 4, and feedback vertex set size 2. However, Bateni, Hajiaghayi and Marx [JACM, 2011] gave an approximation scheme with a runtime of $n^{O(frac{k^2}{varepsilon})}$ on graphs of treewidth $k$. Our main result is a much faster efficient parameterized approximation scheme (EPAS) with a runtime of $2^{O(frac{k^2}{varepsilon} log frac{k^2}{varepsilon})} cdot n^{O(1)}$. If $k$ instead is the vertex cover number of the input graph, we show how to compute the optimum solution in $2^{O(k log k)} cdot n^{O(1)}$ time, and we also prove that this runtime dependence on $k$ is asymptotically best possible, under ETH. Furthermore, if $k$ is the size of a feedback edge set, then we obtain a faster $2^{O(k)} cdot n^{O(1)}$ time algorithm, which again cannot be improved under ETH.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"12 20","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139962706","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-15DOI: 10.48550/arXiv.2402.09812
Jisu Nam, Heesu Kim, DongJae Lee, Siyoon Jin, Seungryong Kim, Seunggyu Chang
The objective of text-to-image (T2I) personalization is to customize a diffusion model to a user-provided reference concept, generating diverse images of the concept aligned with the target prompts. Conventional methods representing the reference concepts using unique text embeddings often fail to accurately mimic the appearance of the reference. To address this, one solution may be explicitly conditioning the reference images into the target denoising process, known as key-value replacement. However, prior works are constrained to local editing since they disrupt the structure path of the pre-trained T2I model. To overcome this, we propose a novel plug-in method, called DreamMatcher, which reformulates T2I personalization as semantic matching. Specifically, DreamMatcher replaces the target values with reference values aligned by semantic matching, while leaving the structure path unchanged to preserve the versatile capability of pre-trained T2I models for generating diverse structures. We also introduce a semantic-consistent masking strategy to isolate the personalized concept from irrelevant regions introduced by the target prompts. Compatible with existing T2I models, DreamMatcher shows significant improvements in complex scenarios. Intensive analyses demonstrate the effectiveness of our approach.
{"title":"DreamMatcher: Appearance Matching Self-Attention for Semantically-Consistent Text-to-Image Personalization","authors":"Jisu Nam, Heesu Kim, DongJae Lee, Siyoon Jin, Seungryong Kim, Seunggyu Chang","doi":"10.48550/arXiv.2402.09812","DOIUrl":"https://doi.org/10.48550/arXiv.2402.09812","url":null,"abstract":"The objective of text-to-image (T2I) personalization is to customize a diffusion model to a user-provided reference concept, generating diverse images of the concept aligned with the target prompts. Conventional methods representing the reference concepts using unique text embeddings often fail to accurately mimic the appearance of the reference. To address this, one solution may be explicitly conditioning the reference images into the target denoising process, known as key-value replacement. However, prior works are constrained to local editing since they disrupt the structure path of the pre-trained T2I model. To overcome this, we propose a novel plug-in method, called DreamMatcher, which reformulates T2I personalization as semantic matching. Specifically, DreamMatcher replaces the target values with reference values aligned by semantic matching, while leaving the structure path unchanged to preserve the versatile capability of pre-trained T2I models for generating diverse structures. We also introduce a semantic-consistent masking strategy to isolate the personalized concept from irrelevant regions introduced by the target prompts. Compatible with existing T2I models, DreamMatcher shows significant improvements in complex scenarios. Intensive analyses demonstrate the effectiveness of our approach.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"7 9","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139962723","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The extraction of cyber threat intelligence (CTI) from open sources is a rapidly expanding defensive strategy that enhances the resilience of both Information Technology (IT) and Operational Technology (OT) environments against large-scale cyber-attacks. While previous research has focused on improving individual components of the extraction process, the community lacks open-source platforms for deploying streaming CTI data pipelines in the wild. To address this gap, the study describes the implementation of an efficient and well-performing platform capable of processing compute-intensive data pipelines based on the cloud computing paradigm for real-time detection, collecting, and sharing CTI from different online sources. We developed a prototype platform (TSTEM), a containerized microservice architecture that uses Tweepy, Scrapy, Terraform, ELK, Kafka, and MLOps to autonomously search, extract, and index IOCs in the wild. Moreover, the provisioning, monitoring, and management of the TSTEM platform are achieved through infrastructure as a code (IaC). Custom focus crawlers collect web content, which is then processed by a first-level classifier to identify potential indicators of compromise (IOCs). If deemed relevant, the content advances to a second level of extraction for further examination. Throughout this process, state-of-the-art NLP models are utilized for classification and entity extraction, enhancing the overall IOC extraction methodology. Our experimental results indicate that these models exhibit high accuracy (exceeding 98%) in the classification and extraction tasks, achieving this performance within a time frame of less than a minute. The effectiveness of our system can be attributed to a finely-tuned IOC extraction method that operates at multiple stages, ensuring precise identification of relevant information with low false positives.
从开放源中提取网络威胁情报（CTI）是一种快速发展的防御策略，可增强信息技术（IT）和操作技术（OT）环境抵御大规模网络攻击的能力。虽然以前的研究主要集中在改进提取过程的各个组件，但社区缺乏在野外部署流 CTI 数据管道的开源平台。为了弥补这一不足，本研究介绍了一个高效且性能良好的平台的实施情况，该平台能够处理基于云计算范式的计算密集型数据管道，用于实时检测、收集和共享来自不同在线来源的 CTI。我们开发了一个原型平台（TSTEM），它是一个容器化的微服务架构，使用 Tweepy、Scrapy、Terraform、ELK、Kafka 和 MLOps 在野外自主搜索、提取和索引 IOC。此外，TSTEM 平台的配置、监控和管理都是通过基础设施即代码（IaC）实现的。自定义焦点爬虫收集网络内容，然后由一级分类器进行处理，以识别潜在的入侵指标（IOC）。如果认为相关，内容将进入第二级提取，以便进一步检查。在整个过程中，最先进的 NLP 模型被用于分类和实体提取，从而增强了整个 IOC 提取方法。我们的实验结果表明，这些模型在分类和提取任务中表现出了很高的准确率（超过 98%），并在不到一分钟的时间内实现了这一性能。我们系统的有效性可归功于经过微调的 IOC 提取方法，该方法可在多个阶段运行，确保以较低的误报率精确识别相关信息。
{"title":"TSTEM: A Cognitive Platform for Collecting Cyber Threat Intelligence in the Wild","authors":"Prasasthy Balasubramanian, Sadaf Nazari, Danial Khosh Kholgh, A. Mahmoodi, Justin Seby, Panos Kostakos","doi":"10.48550/arXiv.2402.09973","DOIUrl":"https://doi.org/10.48550/arXiv.2402.09973","url":null,"abstract":"The extraction of cyber threat intelligence (CTI) from open sources is a rapidly expanding defensive strategy that enhances the resilience of both Information Technology (IT) and Operational Technology (OT) environments against large-scale cyber-attacks. While previous research has focused on improving individual components of the extraction process, the community lacks open-source platforms for deploying streaming CTI data pipelines in the wild. To address this gap, the study describes the implementation of an efficient and well-performing platform capable of processing compute-intensive data pipelines based on the cloud computing paradigm for real-time detection, collecting, and sharing CTI from different online sources. We developed a prototype platform (TSTEM), a containerized microservice architecture that uses Tweepy, Scrapy, Terraform, ELK, Kafka, and MLOps to autonomously search, extract, and index IOCs in the wild. Moreover, the provisioning, monitoring, and management of the TSTEM platform are achieved through infrastructure as a code (IaC). Custom focus crawlers collect web content, which is then processed by a first-level classifier to identify potential indicators of compromise (IOCs). If deemed relevant, the content advances to a second level of extraction for further examination. Throughout this process, state-of-the-art NLP models are utilized for classification and entity extraction, enhancing the overall IOC extraction methodology. Our experimental results indicate that these models exhibit high accuracy (exceeding 98%) in the classification and extraction tasks, achieving this performance within a time frame of less than a minute. The effectiveness of our system can be attributed to a finely-tuned IOC extraction method that operates at multiple stages, ensuring precise identification of relevant information with low false positives.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"6 12","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139962731","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Large Language Models (LLMs), with their advanced contextual understanding abilities, have demonstrated considerable potential in enhancing recommendation systems via fine-tuning methods. However, fine-tuning requires users' behavior data, which poses considerable privacy risks due to the incorporation of sensitive user information. The unintended disclosure of such data could infringe upon data protection laws and give rise to ethical issues. To mitigate these privacy issues, Federated Learning for Recommendation (Fed4Rec) has emerged as a promising approach. Nevertheless, applying Fed4Rec to LLM-based recommendation presents two main challenges: first, an increase in the imbalance of performance across clients, affecting the system's efficiency over time, and second, a high demand on clients' computational and storage resources for local training and inference of LLMs. To address these challenges, we introduce a Privacy-Preserving LLM-based Recommendation (PPLR) framework. The PPLR framework employs two primary strategies. First, it implements a dynamic balance strategy, which involves the design of dynamic parameter aggregation and adjustment of learning speed for different clients during the training phase, to ensure relatively balanced performance across all clients. Second, PPLR adopts a flexible storage strategy, selectively retaining certain sensitive layers of the language model on the client side while offloading non-sensitive layers to the server. This approach aims to preserve user privacy while efficiently saving computational and storage resources. Experimental results demonstrate that PPLR not only achieves a balanced performance among clients but also enhances overall system performance in a manner that is both computationally and storage-efficient, while effectively protecting user privacy.
{"title":"LLM-based Federated Recommendation","authors":"Jujia Zhao, Wenjie Wang, Chen Xu, Zhaochun Ren, See-kiong Ng, Tat-seng Chua","doi":"10.48550/arXiv.2402.09959","DOIUrl":"https://doi.org/10.48550/arXiv.2402.09959","url":null,"abstract":"Large Language Models (LLMs), with their advanced contextual understanding abilities, have demonstrated considerable potential in enhancing recommendation systems via fine-tuning methods. However, fine-tuning requires users' behavior data, which poses considerable privacy risks due to the incorporation of sensitive user information. The unintended disclosure of such data could infringe upon data protection laws and give rise to ethical issues. To mitigate these privacy issues, Federated Learning for Recommendation (Fed4Rec) has emerged as a promising approach. Nevertheless, applying Fed4Rec to LLM-based recommendation presents two main challenges: first, an increase in the imbalance of performance across clients, affecting the system's efficiency over time, and second, a high demand on clients' computational and storage resources for local training and inference of LLMs. To address these challenges, we introduce a Privacy-Preserving LLM-based Recommendation (PPLR) framework. The PPLR framework employs two primary strategies. First, it implements a dynamic balance strategy, which involves the design of dynamic parameter aggregation and adjustment of learning speed for different clients during the training phase, to ensure relatively balanced performance across all clients. Second, PPLR adopts a flexible storage strategy, selectively retaining certain sensitive layers of the language model on the client side while offloading non-sensitive layers to the server. This approach aims to preserve user privacy while efficiently saving computational and storage resources. Experimental results demonstrate that PPLR not only achieves a balanced performance among clients but also enhances overall system performance in a manner that is both computationally and storage-efficient, while effectively protecting user privacy.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"2 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139962743","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-15DOI: 10.48550/arXiv.2402.10208
Eliahu Horwitz, Jonathan Kahana, Yedid Hoshen
The dominant paradigm in generative modeling consists of two steps: i) pre-training on a large-scale but unsafe dataset, ii) aligning the pre-trained model with human values via fine-tuning. This practice is considered safe, as no current method can recover the unsafe, pre-fine-tuning model weights. In this paper, we demonstrate that this assumption is often false. Concretely, we present Spectral DeTuning, a method that can recover the weights of the pre-fine-tuning model using a few low-rank (LoRA) fine-tuned models. In contrast to previous attacks that attempt to recover pre-fine-tuning capabilities, our method aims to recover the exact pre-fine-tuning weights. Our approach exploits this new vulnerability against large-scale models such as a personalized Stable Diffusion and an aligned Mistral.
{"title":"Recovering the Pre-Fine-Tuning Weights of Generative Models","authors":"Eliahu Horwitz, Jonathan Kahana, Yedid Hoshen","doi":"10.48550/arXiv.2402.10208","DOIUrl":"https://doi.org/10.48550/arXiv.2402.10208","url":null,"abstract":"The dominant paradigm in generative modeling consists of two steps: i) pre-training on a large-scale but unsafe dataset, ii) aligning the pre-trained model with human values via fine-tuning. This practice is considered safe, as no current method can recover the unsafe, pre-fine-tuning model weights. In this paper, we demonstrate that this assumption is often false. Concretely, we present Spectral DeTuning, a method that can recover the weights of the pre-fine-tuning model using a few low-rank (LoRA) fine-tuned models. In contrast to previous attacks that attempt to recover pre-fine-tuning capabilities, our method aims to recover the exact pre-fine-tuning weights. Our approach exploits this new vulnerability against large-scale models such as a personalized Stable Diffusion and an aligned Mistral.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"31 2","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139962752","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-15DOI: 10.48550/arXiv.2402.10192
Philip A. LeMaitre, Marius Krumm, H. Briegel
With the impressive progress of deep learning, applications relying on machine learning are increasingly being integrated into daily life. However, most deep learning models have an opaque, oracle-like nature making it difficult to interpret and understand their decisions. This problem led to the development of the field known as eXplainable Artificial Intelligence (XAI). One method in this field known as Projective Simulation (PS) models a chain-of-thought as a random walk of a particle on a graph with vertices that have concepts attached to them. While this description has various benefits, including the possibility of quantization, it cannot be naturally used to model thoughts that combine several concepts simultaneously. To overcome this limitation, we introduce Multi-Excitation Projective Simulation (mePS), a generalization that considers a chain-of-thought to be a random walk of several particles on a hypergraph. A definition for a dynamic hypergraph is put forward to describe the agent's training history along with applications to AI and hypergraph visualization. An inductive bias inspired by the remarkably successful few-body interaction models used in quantum many-body physics is formalized for our classical mePS framework and employed to tackle the exponential complexity associated with naive implementations of hypergraphs. We prove that our inductive bias reduces the complexity from exponential to polynomial, with the exponent representing the cutoff on how many particles can interact. We numerically apply our method to two toy environments and a more complex scenario modelling the diagnosis of a broken computer. These environments demonstrate the resource savings provided by an appropriate choice of inductive bias, as well as showcasing aspects of interpretability. A quantum model for mePS is also briefly outlined and some future directions for it are discussed.
{"title":"Multi-Excitation Projective Simulation with a Many-Body Physics Inspired Inductive Bias","authors":"Philip A. LeMaitre, Marius Krumm, H. Briegel","doi":"10.48550/arXiv.2402.10192","DOIUrl":"https://doi.org/10.48550/arXiv.2402.10192","url":null,"abstract":"With the impressive progress of deep learning, applications relying on machine learning are increasingly being integrated into daily life. However, most deep learning models have an opaque, oracle-like nature making it difficult to interpret and understand their decisions. This problem led to the development of the field known as eXplainable Artificial Intelligence (XAI). One method in this field known as Projective Simulation (PS) models a chain-of-thought as a random walk of a particle on a graph with vertices that have concepts attached to them. While this description has various benefits, including the possibility of quantization, it cannot be naturally used to model thoughts that combine several concepts simultaneously. To overcome this limitation, we introduce Multi-Excitation Projective Simulation (mePS), a generalization that considers a chain-of-thought to be a random walk of several particles on a hypergraph. A definition for a dynamic hypergraph is put forward to describe the agent's training history along with applications to AI and hypergraph visualization. An inductive bias inspired by the remarkably successful few-body interaction models used in quantum many-body physics is formalized for our classical mePS framework and employed to tackle the exponential complexity associated with naive implementations of hypergraphs. We prove that our inductive bias reduces the complexity from exponential to polynomial, with the exponent representing the cutoff on how many particles can interact. We numerically apply our method to two toy environments and a more complex scenario modelling the diagnosis of a broken computer. These environments demonstrate the resource savings provided by an appropriate choice of inductive bias, as well as showcasing aspects of interpretability. A quantum model for mePS is also briefly outlined and some future directions for it are discussed.","PeriodicalId":8425,"journal":{"name":"ArXiv","volume":"16 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139962991","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：