ACM Computing Surveys最新文献_第3页

Mixed-Precision Quantization for Language Models: Techniques and Prospects 语言模型的混合精度量化：技术与展望

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-16 DOI: 10.1145/3848510

Mariam Rakka, Marios Fournarakis, Olga Krestinskaya, Jinane Bazzi, Khaled Salama, Fadi Kurdahi, Ahmed Eltawil, Mohammed Fouda

{"title":"Mixed-Precision Quantization for Language Models: Techniques and Prospects","authors":"Mariam Rakka, Marios Fournarakis, Olga Krestinskaya, Jinane Bazzi, Khaled Salama, Fadi Kurdahi, Ahmed Eltawil, Mohammed Fouda","doi":"10.1145/3848510","DOIUrl":"https://doi.org/10.1145/3848510","url":null,"abstract":"The rapid scaling of Language Models (LMs) has resulted in unprecedented computational, memory, and energy requirements, making their training and deployment increasingly unsustainable. Quantization has emerged as a crucial compression technique for reducing model size, alleviating memory bottlenecks, and accelerating inference. However, while uniform low-bit quantization (e.g., INT8, INT4) provides significant efficiency gains, it can degrade accuracy in sensitive components of transformer-based LMs. Mixed-precision quantization offers a promising alternative by selectively allocating precision across layers or within tensors to strike a balance between efficiency and accuracy. This survey provides a comprehensive overview of Mixed-Precision quantization frameworks for LMs (MXPLMs). We first review quantization fundamentals, including uniform and non-uniform quantizers, quantization granularity, and methods widely used in post-training quantization. We then categorize and compare recent MXPLM frameworks according to their bit allocation strategies and precision configurations across weights, activations, and key-value caches. A comparative analysis highlights differences in perplexity, zero-shot task performance, and deployment trade-offs. Furthermore, we contrast MXPLMs with earlier mixed-precision quantization methods for deep neural networks, identifying strategies that transfer and those that face challenges in the LM setting. Then, we discuss quantization-compatible hardware, KV cache and mixture-of-experts quantization, and summarize open issues and future directions, including hardware-aware design, activation quantization, and scalable optimization methods for billion-parameter models.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"47 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2026-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148979645","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Towards Optimal Speculative Decoding: A Comprehensive Survey on Optimizations and Future Directions 走向最优的思辨解码：关于优化和未来方向的综合调查

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-16 DOI: 10.1145/3846171

Yuyang Ji, Wenjian Xu, Xiaohong Qian, Lei Zhang, Jian Wan

引用次数: 0

Deep Learning-Based Sign Language Translation: A Survey and Taxonomy 基于深度学习的手语翻译：综述与分类

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-15 DOI: 10.1145/3844609

Desen Yuan, Ruili Wang

引用次数: 0

Survey on Verifiable Machine Learning 可验证机器学习研究综述

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-15 DOI: 10.1145/3847111

He Lyu, Xikun Jiang, Boris Düdder, Yuya Sasaki

引用次数: 0

Beyond Efficiency: A Systematic Survey of Resource-Efficient Large Language Models 超越效率：资源效率型大型语言模型的系统研究

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-15 DOI: 10.1145/3845797

Guangji Bai, Zheng Chai, Chen Ling, Shiyu Wang, Jiaying Lu, Nan Zhang, Tingwei Shi, Ziyang Yu, Mengdan Zhu, Yifei Zhang, Xinyuan Song, Carl Yang, Yue Cheng, Liang Zhao

{"title":"Beyond Efficiency: A Systematic Survey of Resource-Efficient Large Language Models","authors":"Guangji Bai, Zheng Chai, Chen Ling, Shiyu Wang, Jiaying Lu, Nan Zhang, Tingwei Shi, Ziyang Yu, Mengdan Zhu, Yifei Zhang, Xinyuan Song, Carl Yang, Yue Cheng, Liang Zhao","doi":"10.1145/3845797","DOIUrl":"https://doi.org/10.1145/3845797","url":null,"abstract":"The burgeoning field of Large Language Models (LLMs), exemplified by sophisticated models like OpenAI’s ChatGPT, represents a significant advancement in artificial intelligence. These models, however, bring forth substantial challenges in high consumption of computational, memory, energy, and financial resources, especially in environments with limited resource capabilities. This survey aims to systematically address these challenges by reviewing a broad spectrum of techniques designed to enhance the resource efficiency of LLMs. We categorize methods based on their optimization focus—covering computational, memory, energy, financial, and network resources—and their applicability across various stages of an LLM’s lifecycle, including architecture design, pre-training, fine-tuning, and system design. Additionally, the survey introduces a nuanced categorization of resource efficiency techniques by their specific resource types, which uncovers the intricate relationships and mappings between various resources and corresponding optimization techniques. A standardized set of evaluation metrics and datasets is also presented to facilitate consistent and fair comparisons across different models and techniques. By offering a comprehensive overview of the current state-of-the-art and identifying open research avenues, this survey serves as a foundational reference for researchers and practitioners, aiding them in developing more sustainable and efficient LLMs in a rapidly evolving landscape. To make the field more accessible to interested beginners, we not only make a systematic review of existing works and a highly structured taxonomy of resource-efficient LLMs but also release a website including a constantly-updated paper list https://github.com/tiingweii-shii/Awesome-Resource-Efficient-LLM-Papers.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"7 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2026-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148964304","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Systematic Review of Modern Machine Learning in Query Optimization 现代机器学习在查询优化中的系统综述

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-11 DOI: 10.1145/3847509

Emmanouil Limnaios, Haridimos Kondylakis

引用次数: 0

A Survey on the Evolution and Future Trajectory of Flow Cytometry Analysis 流式细胞术分析的发展和未来发展趋势综述

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-11 DOI: 10.1145/3845984

Tony Xu, Givanna Putri, Aaron Chuah, Robin Vlieger, Anne Brüstle

{"title":"A Survey on the Evolution and Future Trajectory of Flow Cytometry Analysis","authors":"Tony Xu, Givanna Putri, Aaron Chuah, Robin Vlieger, Anne Brüstle","doi":"10.1145/3845984","DOIUrl":"https://doi.org/10.1145/3845984","url":null,"abstract":"Flow cytometry is a powerful analytical technique that generates high-dimensional single-cell data widely used in biological research and clinical diagnostics. This survey traces the evolution of flow cytometry analysis from traditional manual gating to modern algorithmic approaches and explores emerging deep-learning (DL) paradigms. We delineate three distinct analysis paradigms: traditional (manual gating), modern (algorithm-assisted population identification), and future (direct cell-to-sample DL models). While the modern paradigm employs sophisticated computational methods to identify cell populations and derive features, the emerging future paradigm bypasses discrete population identification entirely, utilizing DL to directly connect cellular information to sample-level insights. We discuss key technical challenges including batch effects, panel variations, and data availability that currently limit widespread adoption of more advanced analytical approaches. The survey also examines promising solutions such as transfer learning, data augmentation, and self-supervised learning techniques. As flow cytometry technology continues to advance, particularly with spectral cytometry enabling higher-dimensional analysis, these computational methods will become increasingly essential for extracting maximum biological insight from complex datasets. The development of standardized data repositories, robust batch correction tools, and comprehensive benchmarking frameworks will be crucial for realizing the full potential of these advanced analytical approaches.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"2 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2026-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148922647","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Deep Learning of EEG Signals for Brain Function and Injury: A Systematic Review 脑电信号深度学习对脑功能和损伤的研究综述

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-11 DOI: 10.1145/3847656

Tiago Souza, Murillo Carneiro, Zhao Liang, Yaochu Jin

{"title":"Deep Learning of EEG Signals for Brain Function and Injury: A Systematic Review","authors":"Tiago Souza, Murillo Carneiro, Zhao Liang, Yaochu Jin","doi":"10.1145/3847656","DOIUrl":"https://doi.org/10.1145/3847656","url":null,"abstract":"Electroencephalography (EEG) is a valuable, low-cost tool for capturing brain signals, typically through non-invasive scalp recordings. However, analyzing EEG signals poses challenges due to noise interference and the need for specialized professionals. Recent advancements in deep learning have addressed many of these issues, leading to significant progress. Despite numerous systematic reviews, there remains a gap in deep learning research focused on assessing consciousness, brain injury, and comatose states. To bridge this gap, we conducted a systematic literature review following the PRISMA protocol, targeting human studies published between 2018 and 2023 across various databases. We identified 475 studies, of which 37 met our eligibility criteria. Their analyses revealed key trends: i) most focus on classification tasks, particularly for consciousness assessment and seizure detection; ii) most datasets involve fewer than 100 participants, with few being publicly available; iii) Convolutional Neural Networks dominate the learning architecture, although emerging approaches like Graph Attention Networks and Transformers show promising performance. However, challenges remain in distributed data acquisition, data augmentation, automated and multimodal machine learning, hyperparameter optimization, and model explainability, which are underexplored. In conclusion, while deep learning has advanced EEG signal analysis, further research is essential to build larger datasets and improve multimodal models for broader applications in Brain Function and Injury assessment.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"212 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2026-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148922648","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Exploring Backdoor Vulnerabilities and Fairness Challenges in Generative Models: A Survey 探索生成模型中的后门漏洞和公平挑战：综述

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-11 DOI: 10.1145/3847107

Ryan Holland, Shantanu Pal, Lei Pan

{"title":"Exploring Backdoor Vulnerabilities and Fairness Challenges in Generative Models: A Survey","authors":"Ryan Holland, Shantanu Pal, Lei Pan","doi":"10.1145/3847107","DOIUrl":"https://doi.org/10.1145/3847107","url":null,"abstract":"Global connectivity has increased societal diversity worldwide, fostering growth and dispelling negative stereotypes. However, ensuring equality remains challenging, with some groups favoured and others facing unequal opportunities. The evolution of Artificial Intelligence (AI) promises automation, innovation, and industry transformation. Yet, as AI becomes pervasive, establishing trust in algorithmic decision-making is crucial to eliminate bias and ensure fair outcomes. Adversarial attacks, particularly backdoor attacks, pose significant threats to AI systems. Research must focus on mitigating discrimination and bias in AI models, emphasising ethical considerations and model interpretability. Accordingly, this paper surveys the critical aspects of fairness, bias, and backdoor attacks in AI systems, integrating significant and state-of-the-art research. This paper begins by defining key concepts, including fairness, bias, and backdoor attacks, and exploring related elements. A thorough survey highlights key studies on fairness and bias analysis, metrics for measurement, and various fairness attacks. This paper further examines methodologies for performing backdoor attacks and their implications for AI systems. Additionally, this paper explores the latest defence mechanisms addressing fairness and bias issues and mitigating backdoor attacks, showcasing advancements in safeguarding AI integrity. The survey concludes with discussions on popular datasets, future research directions, and potential limitations in ongoing developments. This is the first study combining surveys of backdoor attacks and fairness in AI models to examine security vulnerabilities that can lead to bias susceptibility and unfair outcomes. By consolidating crucial research insights, this survey provides a comprehensive understanding of the challenges and progress in ensuring fairness and security in AI systems, offering valuable perspectives for researchers and practitioners.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"52 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2026-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148922649","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Survey on Foundation Models for Personalized Federated Intelligence 个性化联邦智能基础模型研究

IF 16.6 1区计算机科学

ACM Computing Surveys Pub Date : 2026-09-09 DOI: 10.1145/3844947

Yu Qiao, Huy Le, Avi Raha, Phuong-Nam Tran, Apurba Adhikary, Mengchun Zhang, Loc X. Nguyen, Eui-Nam Huh, Dusit (Tao) Niyato, Choong Seon Hong

{"title":"A Survey on Foundation Models for Personalized Federated Intelligence","authors":"Yu Qiao, Huy Le, Avi Raha, Phuong-Nam Tran, Apurba Adhikary, Mengchun Zhang, Loc X. Nguyen, Eui-Nam Huh, Dusit (Tao) Niyato, Choong Seon Hong","doi":"10.1145/3844947","DOIUrl":"https://doi.org/10.1145/3844947","url":null,"abstract":"The rise of large language models (LLMs), such as ChatGPT, Gemini, and Grok, has reshaped the AI landscape. As prominent instances of foundational models (FMs), they exhibit remarkable capabilities in generating human-like content, pushing the boundaries towards artificial general intelligence (AGI). However, their large-scale nature, privacy sensitivity, and substantial computational demands pose significant challenges for personalized customization for end users. To bridge this gap, we present the vision of artificial personalized intelligence (API), which focuses on adapting FMs to individual users while ensuring privacy. As a central enabler of API, we propose personalized federated intelligence (PFI), a new paradigm that not only integrates the privacy benefits of federated learning (FL) with the generalization capabilities of FMs but also places personalization at its core. To this end, we first survey recent advances in FL and FMs that lay the foundation for PFI. We then explore core stages of the PFI pipeline: efficient personalization at the edge, trustworthy adaptation, and adaptive refinement via retrieval-augmented generation. Finally, we highlight future directions for enabling PFI. Overall, this survey aims to lay a foundation for the development of API as a complementary direction to AGI, with PFI as a key enabling paradigm.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"16 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2026-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148909194","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0