🧑‍🔬 Ion Gabriel Stoica

🏛️ Affiliations

🎓University of California, BerkeleyBerkeley · United States2000–2026 · 442 papersEmployment since 2000 · Professor, Electrical Engineering and Computer Sciencecurrent🏢OranoWashington D.C. · United States2026 · 1 papercurrent🎓University of California, Santa CruzSanta Cruz · United States2026 · 1 papercurrent🎓Carnegie Mellon UniversityPittsburgh · United States1997–2003 · 16 papers🎓Old Dominion UniversityNorfolk · United States1994–2002 · 14 papers🏢DatabricksSan Francisco2014–2018 · 7 papers🔬International Computer Science InstituteBerkeley · United States2003–2024 · 4 papers🔬Central University LibraryBucharest · Romania1992–1995 · 4 papers🎓Stanford UniversityStanford · United States2024 · 2 papers🎓Berkeley CollegeNew York · United States2017–2019 · 2 papers

From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 18 more institutions on single bylines are not listed.

🧭 Fields, topics and concepts

Of its works in our library, with the number of works on each.

📅 Works by year

2020s

📄 Papers

The most cited first (OpenAlex and our library).

1. Efficient Memory Management for Large Language Model Serving with PagedAttention (2023 · 4,531 citations)
   Serving large language models efficiently depends on how many requests a GPU can process together. The limiting factor is often the KV cache: the stored attention keys and values for each request, whose size changes as tokens are generated and whose final length is unknown in advance. Earlier systems such as…
2. SGLang: Efficient Execution of Structured Language Model Programs (2023 · 729 citations)
   Modern LLM applications such as agents, few-shot evaluation, tree-of-thought search, RAG and JSON extraction rarely make a single model call.
3. LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code (2024 · 578 citations)
   Widely used code benchmarks such as HumanEval and MBPP have become saturated. They also risk leaking into training data, and they test only one skill: turning a natural-language description into code. LiveCodeBench addresses these problems with a benchmark that keeps growing. It collects fresh problems from weekly…
4. Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference (2024 · 546 citations)
   The authors show that conventional LLM benchmarks poorly reflect how models are actually used. These are fixed sets of closed-ended questions with reference answers: over time they leak into training data, become saturated, and do not account for human preferences in open-ended tasks. As an alternative, they propose…
5. LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset (2023 · 242 citations)
   Research into how people use large language models has been held back by the lack of public logs of genuine user interactions.
6. From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipeline (2024 · 210 citations)
   Static LLM benchmarks saturate quickly, can leak into training data, and mostly test close-ended tasks. Expert-built sets are expensive: GPQA's 500 questions reportedly cost over $120,000. Live human evaluation such as Chatbot Arena is too slow for developers to use routinely. The authors present BenchBuilder, an…
7. RAFT: Adapting Language Model to Domain Specific RAG (2024 · 137 citations)
   The paper asks how a pretrained LLM should be prepared when it will answer questions over a known, fixed document collection, such as enterprise files, medical literature or API documentation, with retrieval at inference time.
8. Break the Sequential Dependency of LLM Inference Using Lookahead Decoding (2024 · 114 citations)
   Generating text with LLMs one token at a time is slow. Each step is limited by memory bandwidth, so most of the GPU's compute sits idle. Speculative decoding helps, but it needs a draft model, which is hard to train and often fails to transfer across base models or datasets. The authors propose Lookahead Decoding…
9. Rethinking Benchmark and Contamination for Language Models with Rephrased Samples (2023 · 103 citations)
   Public LLM benchmarks can leak into training corpora. Standard decontamination looks for surface similarity, using either n-gram overlap or an embedding-similarity threshold. The authors show that these checks are easy to get around. They had GPT-4 reword test items, or rewrite code in a different style or…
10. OR-Bench: An Over-Refusal Benchmark for Large Language Models (2024 · 76 citations)
    Safety-tuned LLMs often refuse harmless requests that merely look risky. Before this work, the problem had no large-scale way to be measured. The only dedicated test set, XSTest, has 250 hand-written prompts, and recent strong models now answer almost all of them. The authors build an automated pipeline to fill this…

All works

584 works of theirs in our library, the most cited first; 10 with a page here, the others link their original until we summarise them.

1. Chord ↗ACM SIGCOMM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications (SIGCOMM) · no summary yetIon Gabriel Stoica et al.2001 · 9,623 citations
2. Above the Clouds: A Berkeley View of Cloud Computing ↗no summary yetMichael Armbrust et al.2009 · 5,696 citations
3. Efficient Memory Management for Large Language Model Serving with PagedAttentionarXiv:2309.06180Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, et al.2023 · 4,531 citations
4. Spark: cluster computing with working sets ↗no summary yetMatei Zaharia et al.2010 · 4,118 citations
5. Resilient distributed datasets: a fault-tolerant abstraction for in-memory cluster computing ↗no summary yetMatei Zaharia et al.2012 · 3,546 citations
6. Chord: a scalable peer-to-peer lookup protocol for internet applications ↗IEEE/ACM Transactions on Networking · no summary yetIon Gabriel Stoica et al.2002 · 3,135 citations
7. Apache Spark ↗Communications of the ACM · no summary yetMatei Zaharia et al.2016 · 2,436 citations
8. Chord ↗ACM SIGCOMM Computer Communication Review · no summary yetIon Gabriel Stoica et al.2001 · 1,923 citations
9. Improving MapReduce performance in heterogeneous environments ↗no summary yetMatei Zaharia et al.2008 · 1,592 citations
10. Mesos: a platform for fine-grained resource sharing in the data center ↗no summary yetBenjamin Hindman et al.2011 · 1,582 citations
11. Wide-area cooperative storage with CFS ↗no summary yetFrank Dabek et al.2001 · 1,430 citations
12. Delay scheduling ↗European Conference on Computer Systems · no summary yetMatei Zaharia et al.2010 · 1,395 citations
13. A data-oriented (and beyond) network architecture ↗ACM SIGCOMM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications (SIGCOMM) · no summary yetTeemu Koponen et al.2007 · 1,368 citations
14. Dominant resource fairness: fair allocation of multiple resource types ↗no summary yetAli Ghodsi et al.2011 · 1,077 citations
15. Geographic routing without location information ↗Proceedings of the 28th Annual International Conference on Mobile Computing And Networking · no summary yetAnanth Rao et al.2003 · 748 citations
16. SGLang: Efficient Execution of Structured Language Model ProgramsConference on Neural Information Processing Systems (NeurIPS)Lianmin Zheng, Liangsheng Yin, Zhiqiang Xie, et al.2023 · 729 citations
17. BlinkDB ↗no summary yetSameer Agarwal et al.2013 · 694 citations
18. Understanding the impact of video quality on user engagement ↗ACM SIGCOMM Conference (SIGCOMM) · no summary yetFlorin Dobrian et al.2011 · 674 citations
19. Internet indirection infrastructure ↗ACM SIGCOMM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications (SIGCOMM) · no summary yetIon Gabriel Stoica et al.2002 · 642 citations
20. Tune: A Research Platform for Distributed Model Selection and Training ↗arXiv (Cornell University) · no summary yetRichard Liaw et al.2018 · 640 citations
21. Reining in the outliers in map-reduce clusters using Mantri ↗Operating Systems Design and Implementation · no summary yetGanesh Ananthanarayanan et al.2010 · 639 citations
22. GraphX ↗no summary yetReynold Xin et al.2013 · 581 citations
23. LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for CodearXiv:2403.07974Naman Jain, King Han, Alex Gu, et al.2024 · 578 citations
24. The impact of DHT routing geometry on resilience and proximity ↗ACM SIGCOMM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications (SIGCOMM) · no summary yetKrishna P. Gummadi et al.2003 · 567 citations
25. Robust incentive techniques for peer-to-peer networks ↗ACM Conference on Electronic Commerce (EC) · no summary yetMichal Feldman et al.2004 · 558 citations

See all 584 works by Ion Gabriel Stoica →

🤝 Co-authors here

12 of 30 co-authors with a page, the most papers together first.

• Joseph E. Gonzalez ×7
• Lianmin Zheng ×5
• Wei-Lin Chiang ×5
• Ying Sheng ×4
• Hao Zhang ×3
• Tianle Li ×3
• Banghua Zhu ×2
• Cody Hao Yu ×2
• Naman Jain ×2
• Siyuan Zhuang ×2
• Tianjun Zhang ×2
• Zhuohan Li ×2

Venues

Where their works here were published, the most works first, with the editions of those that have a page.

• Lecture notes in computer science ×26
• PubMed ×26
• Proceedings of the VLDB Endowment ×18
• ACM SIGCOMM Computer Communication Review ×14
• Proceedings - ACM-SIGMOD International Conference on Management of Data ×12
• Networked Systems Design and Implementation ×11
• neural information processing systems ×11
• ACM SIGCOMM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications (SIGCOMM) ×8
• UC Berkeley ×8
• ACM Symposium on Cloud Computing (SoCC) ×5
• ACM Workshop on Hot Topics in Networks (HotNets) ×5
• IEEE/ACM Transactions on Networking ×5

🔗 Cited by

Papers of our library cite their works 8,499 times (of the 149,439 papers whose bibliographies we read); 859 citing papers with a page here →.

Updated 30 October 2026

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/authors/ion-gabriel-stoica/