April 2024 - talks - lists.cs.princeton.edu

CS Colloquium Speakers: week of March 4
by Emily C. Lawrence 03 Dec '24

03 Dec '24

CS Colloquium Speakers Speaker: Sewon Min, University of Washington Date: Monday, March 4 Time: 12:30pm EST Location: CS 105 Host: Danqi Chen Event page: https://www.cs.princeton.edu/events/26579 Register for live-stream online here: https://princeton.zoom.us/webinar/register/WN_5uVGEpXNTVe8eBNSLpxKtg Title: Rethinking Data Use in Large Language Models Abstract: Large language models (LMs) such as ChatGPT have revolutionized natural language processing and artificial intelligence more broadly. In this talk, I will discuss my research on understanding and advancing these models, centered around how they use the very large text corpora they are trained on. First, I will describe our efforts to understand how these models learn to perform new tasks after training, demonstrating that their so-called in context learning capabilities are almost entirely determined by what they learn from the training data. Next, I will introduce a new class of LMs—nonparametric LMs—that repurpose this training data as a data store from which they retrieve information for improved accuracy and updatability. I will describe my work on establishing the foundations of such models, including one of the first broadly used neural retrieval models and an approach that simplifies a traditional, two-stage pipeline into one. I will also discuss how nonparametric models open up new avenues for responsible data use, e.g., by segregating permissive and copyrighted text and using them differently. Finally, I will envision the next generation of LMs we should build, focusing on efficient scaling, improved factuality, and decentralization. Bio: Sewon Min is a Ph.D. candidate in the Paul G. Allen School of Computer Science & Engineering at the University of Washington. Her research focuses on language models (LMs): studying the science of LMs, and designing new model classes and learning methods that make LMs more performant and flexible. She also studies LMs in information-seeking, legal, and privacy contexts. She is a co-organizer of multiple tutorials and workshops, including most recently at ACL 2023 on Retrieval-based Language Models and Applications and upcoming at ICLR 2024 on Mathematical and Empirical Understanding of Foundation Models. She won a paper award at ACL 2023, received a J.P. Morgan Fellowship, and was named an EECS rising star in 2022. Speaker: Lianmin Zheng, University of California, Berkeley Date: Tuesday, March 5 Time: 12:30pm EST Location: CS 105 Host: Ravi Netravali Event page: https://www.cs.princeton.edu/events/26593 Register for live-stream online here: https://princeton.zoom.us/webinar/register/WN_ghiT2HbhSPqKBouJ-VwzFg Title: Scalable and Efficient Systems for Large Language Models Abstract: Large Language Models (LLMs) have been driving recent breakthroughs in AI. These advancements would not have been possible without the support of scalable and efficient infrastructure systems. In this talk, I will introduce several underlying systems I have designed and built to support the entire model lifecycle, from training to deployment to evaluation. First, I will present Alpa, a system for large-scale model-parallel training that automatically generates execution plans unifying data, operator, and pipeline parallelism. Next, I will discuss efficient deployment systems, covering the frontend programming interface and backend runtime optimizations for high-performance inference. Finally, I will complete the model lifecycle by presenting our model evaluation efforts, including the crowdsourced live benchmark platform, Chatbot Arena, and the automatic evaluation pipeline, LLM-as-a-Judge. These projects have collectively laid a solid foundation for large language model systems, being widely adopted by leading LLM developers and companies. I will conclude by outlining some future directions of machine learning systems, such as co-optimizing across the full stack for building AI-centric applications. Bio: Lianmin Zheng is a Ph.D. student in the EECS department at UC Berkeley, advised by Ion Stoica and Joseph E. Gonzalez. His research interests include machine learning systems, large language models, compilers, and distributed systems. He builds full-stack, scalable, and efficient systems to advance the development of AI. He co-founded LMSYS.org, where he leads impactful open-source large language model projects such as Vicuna and Chatbot Arena, which have received millions of downloads and served millions of users. He also co-organized the Big Model Tutorial at ICML 2022. He has received a Meta Ph.D. Fellowship, an IEEE Micro Best Paper Award, and an a16z open-source AI grant. Speaker: Emma Dauterman, University of California, Berkeley Date: Thursday, March 7 Time: 12:30pm EST Location: CS 105 Host: Amit Levy Event page: https://www.cs.princeton.edu/events/26585 Register for live-stream online here: https://princeton.zoom.us/webinar/register/WN_soNesOKXSFml9MPaTx_1iw Title: Secure systems from insecure components Abstract: In many computer systems today, an attacker that breaks one system component can steal data from millions of users. In this talk, I will present two systems that can withstand component compromise. I will describe (1) a single sign-on system that protects user security and privacy from a compromised single sign-on server, and (2) a secure-hardware-based backup service that protects user backups from compromised secure hardware devices. These systems provide strong security and privacy properties while taking into account practical constraints such as compatibility requirements, hardware limitations, and user expectations. Each splits user secrets across different system components, using new cryptographic tools to provide necessary functionality while protecting user data. Bio: Emma Dauterman is a Ph.D. candidate at UC Berkeley where she is advised by Raluca Ada Popa and Ion Stoica. Her research interests include computer security, systems, and applied cryptography. She has received the Microsoft Research Ada Lovelace fellowship, the NSF graduate research fellowship, and a UC Berkeley EECS excellence award.

1 22

Dmitry Paramonov will present his FPO "Handling Data Too Large To Handle: On Multi-pass Streaming and Interactive Coding" on May 7th, 2024 at 11am in Friend 006.
by Gradinfo 07 May '24

07 May '24

Dmitry Paramonov will present his FPO "Handling Data Too Large To Handle: On Multi-pass Streaming and Interactive Coding" on May 7th, 2024 at 11am in Friend 006. His committee is as follows: Examiners: Gillat Kol (adviser), Ran Raz, Mark Braverman Readers: Huacheng Yu and Matt Weinberg Title: Handling Data Too Large To Handle: On Multi-pass Streaming and Interactive Coding Abstract: Over the last decades, the world has become increasingly more information-centric and massive amounts of data, potentially distributed between many different sources, are being processed all the time. In this thesis, I consider two mechanisms for coping with big data and the distributed nature of timely tasks. In Part I, I showcase my work on the streaming setting, where the input to the algorithm is given as a stream of elements. The algorithm’s goal is to compute a value that depends on the stream while only utilizing memory that is much smaller than the entire stream. My work in this field focuses on proving that various fundamental graph problems essentially require the streaming algorithm to store the entire graph, even if it is allowed to make several passes through the given stream of edges. In Part II, I consider error-correcting codes for distributed, interactive settings. Classical error- correcting codes assume that a sender who has all the information wishes to send it to a receiver over a noisy channel. However, in many modern, big data applications, the information is distributed amongst many parties that communicate back-and-forth to compute a value that depends on all their inputs. My work examines the noise resilience of various such settings. For some models, we can design error-correcting protocols that allow the encoding of every noiseless protocol by a noise-resilient protocol with low overhead, whereas for other models, it can be shown that this task is impossible. While these two topics appear greatly unrelated, and almost orthogonal to one another, the tools used to prove results in both turn out to be remarkably similar, with many standard problems and information theory lemmas being a critical part of both.

1 1

Tamjeed Azad will present his General Exam "Exploring and modeling intercellular interactions and microenvironments." on Friday, May 3, 2024 at 3:00 PM in Icahn 280.
by CS Grad Department 30 Apr '24

30 Apr '24

Tamjeed Azad will present his General Exam "Exploring and modeling intercellular interactions and microenvironments." on Friday, May 3, 2024 at 3:00 PM in Icahn 280. Committee Members: Yuri Pritykin (advisor), Ben Raphael, Mona Singh Abstract: https://docs.google.com/document/d/1B-6aCqVqQEK5OErdrKlFO-jXmnf69iZZiyd7Zqu… Reading List: https://docs.google.com/document/d/17gQJZYG3smNJZ74ayQhUwP38kYdPTXiFLMlcyTP… Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so.

1 0

Isabel Armour-Garb will present her General Exam "Leveraging Protein Language Models for Protein Characterization" on Friday, May 3, 2024 at 10:00 AM in Icahn 280 .
by CS Grad Department 30 Apr '24

30 Apr '24

Isabel Armour-Garb will present her General Exam "Leveraging Protein Language Models for Protein Characterization" on Friday, May 3, 2024 at 10:00 AM in Icahn 280 Committee Members: Mona Singh (advisor), Olga Troyanskaya, Yuri Pritykin Abstract: Genomes are sequenced much faster than experimental efforts to functionally characterize the proteins they contain. Computational methods based on sequence similarity are the first line of attack for annotating sequences with protein function. However, no functional annotations exist for over 1.1 billion protein sequences, and thousands of human proteins lack sufficient characterization. To understand protein function, it is important to identify functional sites (e.g. amino acids involved in crucial interactions such as catalysis or post-translational modifications), which are highly conserved across homologs. However, for proteins with few known homologs, current methods cannot predict evolutionary conservation, because state-of-the-art methods rely on the alignment of homologs. Current methods align these homologs in a multiple sequence alignment to find evolutionarily conserved positions. Protein language models (PLMs) are advanced deep learning natural language processing models specifically designed to understand and generate protein-related texts and sequences. These models represent each amino acid within a protein sequence as a high-dimensional vector, effectively capturing complex information about the amino acids and their sequence context, and providing a way to encode and process the inherent features of the protein as an embedding for computational analysis. First, we introduce vcMSA, a new method for more accurate multiple sequence alignment using PLMs. Then, leveraging PLMs, we introduce a new method for accurately predicting functional sites given a single protein sequence, which allows us to better characterize proteins of unknown function regardless of the presence or absence of known homologs. Reading List: https://docs.google.com/document/d/1thOoMwpz7H1JrgQKYH9Va-Xf5ZyEQnJV-11VeNB… Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so.

1 0

Qianfan Zhang will present his General Exam "Communication Separations for Truthful Auctions: Breaking the Two-Player Barrier" on May 7, 2024 at 1pm in Friend 005.
by Nicki Mahler 30 Apr '24

30 Apr '24

Qianfan Zhang will present his General Exam " Communication Separations for Truthful Auctions: Breaking the Two-Player Barrier" on May 7, 2024 at 1pm in Friend 005. Committee Members: Matt Weinberg (adviser), Mark Braverman, and Gillat Kol Abstract follows below. Combinatorial auctions are a paradigmatic problem at the intersection of Economics and Computation. However, it is still quite poorly understood how well one can approximate the optimal social welfare via truthful mechanisms. In most settings, it remains unknown whether there should be any separation between truthful mechanisms and non-truthful protocols at all. In this work, we study the communication complexity of truthful combinatorial auctions, and in particular the case where valuations are either subadditive or single-minded, which we denote with SubAdd∪SingleM. We show that for three bidders with valuations in SubAdd∪SingleM, any deterministic truthful mechanism which achieves at least a 0.366-approximation requires exp(m) communication. In contrast, a natural extension of [Feige09] yields a non-truthful poly(m)-communication protocol that achieves a 1/2-approximation, demonstrating a gap between the power of truthful mechanisms and non-truthful protocols for this problem. Our approach follows the taxation complexity framework laid out in [Dobzinski16], but applies this framework in a setting not encompassed by the techniques used in past work. In particular, the only successful prior application of this framework uses a reduction to simultaneous protocols which only applies for two bidders [AKSW20], whereas our three-player lower bounds are stronger than what can possibly arise from a two-player construction (since a trivial truthful auction guarantees a 1/2-approximation for two players). Abstract & Reading List: [ https://docs.google.com/document/d/1YfmRXSLmvQbUx2u1Q9IEeK5d9Y6ymQVTLaEEO5o… | https://docs.google.com/document/d/1YfmRXSLmvQbUx2u1Q9IEeK5d9Y6ymQVTLaEEO5o… ]

1 0

Allison Chen will present her General Exam "Understanding Visual Processing In Vision Language Models" on Thursday, May 2, 2024 at 1:30 PM in CS 302 and via zoom.
by CS Grad Department 29 Apr '24

29 Apr '24

Allison Chen will present her General Exam "Understanding Visual Processing In Vision Language Models" on Thursday, May 2, 2024 at 1:30 PM in CS 302 and via zoom. Zoom link: https://princeton.zoom.us/j/99275578452 Committee Members: Olga Russakovsky (advisor), Tom Griffiths, Szymon Rusinkiewicz Abstract: Applications in modern platforms are complex and thus hard to reason about their security. For instance, modern web applications have largely become cloud-based due to the flexibility and cost efficiency brought by the cloud. However, writing secure web applications is challenging as developers need to devise sufficient security measures to ensure intended policies are enforced while applications are cross-components and rapidly changing during development. To address this problem, we propose Faasten, which is a secure cloud architecture and implementation that enforces end-user data policies and supports general cloud applications with negligible performance overheads. Faasten uses decentralized information flow control (DIFC) and leverages Function-as-a-Service (FaaS) programming models to provide a safe, general, and self-encapsulating cloud interface for application developers. Additionally, another challenge that remains unsolved is timing channels. Faasten is unable to close timing channels in cloud platforms because such channels are a result of resource sharing which is out of control of Faasten. Timing channels are critical not only to cloud platforms but also in critical infrastructure. Existing defenses and mitigations are ad-hoc and often come with large performance penalties. With Leak Avoidant Resource Provisioners (LARPS), we argue that a large class of timing channels can be closed by carefully controlling information flow through resource provisioning decisions. In addition, we require a minimal set of hardware primitives to ensure LARP operations themselves do not impose additional timing channels. Reading List: https://docs.google.com/document/d/1-sXQFoc8srt-hFcFyCmfxvFQlbJPCJA1vCuWKNs… Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so.

2 1

ECE/COS Seminar: Luca Carloni, Thursday, May 2 at 12:30pm
by Donna Ghilino 29 Apr '24

29 Apr '24

ECE/COS Seminar Speaker: Luca Carloni, Columbia University Date: Thursday, May 2,2024 Time: 12:30pm Location: CS 105 Host: Margaret Martonosi & Sharad Malik Event page: https://www.cs.princeton.edu/events/26654 Title: Collaborative System-on-Chip Design with Open-Source Hardware Platforms Abstract: Open-source hardware holds a promise of sustaining the progress of the semiconductor industry in the age of heterogeneous computing. It can enable design reuse, foster design collaboration, and support workforce development. ESP is an open-source research platform for system-on-chip (SoC) design that combines a modular architecture and an agile design methodology. The ESP architecture simplifies the design and prototyping of heterogeneous chips with multiple RISC-V processor cores and dozens of loosely-coupled accelerators, all interconnected with a scalable network-on-chip. The ESP methodology promotes system-level design while accommodating different specification languages and design flows. ESP's capabilities have allowed a small team of mostly graduate students to realize two SoCs of growing complexity, each in the span of just a few months. Conceived as a heterogeneous system integration platform and developed through years of teaching at Columbia University, ESP is intrinsically suited to advance collaborative engineering across the open-source hardware community. Bio: Luca Carloni is professor and chair of Computer Science at Columbia University in the City of New York. He holds a Laurea Degree Summa cum Laude in Electronics Engineering from the University of Bologna, Italy, and the MS and PhD degrees in Electrical Engineering and Computer Sciences from the University of California, Berkeley. His research interests include heterogeneous computing, system-on-chip platforms, embedded systems, and open-source hardware. He co-authored over one hundred and eighty refereed papers. Luca received the NSF CAREER Award, the Alfred P. Sloan Research Fellowship, and the ONR Young Investigator Award. He is an IEEE Fellow.

1 0

Jingyuan Chen will present the General Exam "Lumos: Provenance-guided Automatic Online Debugging" on 5/3/24 at 3pm in CS 105 and Zoom.
by Gradinfo 29 Apr '24

29 Apr '24

Jingyuan Chen will present the General Exam "Lumos: Provenance-guided Automatic Online Debugging" on 5/3/24 at 3pm in CS 105 and Zoom. The zoom link is: [ https://princeton.zoom.us/j/95830334391 | https://princeton.zoom.us/j/95830334391 ] The members of the committee are as follows: Amit Levy (adviser), Ravi Netravali, Mae Milano Abstract: [ https://docs.google.com/document/d/1_yVb8aS7wQNMpnLz0PLAFmemiwnIGNZ1W9tPe4A… | https://docs.google.com/document/d/1_yVb8aS7wQNMpnLz0PLAFmemiwnIGNZ1W9tPe4A… ]

1 1

Aatmik Gupta will present his MSE talk “Selecting LLM Training Data using Subjective Qualities” on Monday, April 29, 2024 at 9:30am in Friend 125.
by Gradinfo 26 Apr '24

26 Apr '24

Aatmik Gupta will present his MSE talk “Selecting LLM Training Data using Subjective Qualities” on Monday, April 29, 2024 at 9:30am in Friend 125. The members of his committee are as follows: Danqi Chen (Adviser) and Karthik Narasimhan (reader) All ae welcome to attend. Please see abstract below. I will discuss a novel method for selecting pre-training data that focuses on capturing abstract text qualities that are intuitively recognized by humans. The focus is on four qualities: writing style, required expertise, facts & trivia, and educational value. We find that LLMs are better at comparing these qualities between texts rather than directly assessing a text's quality. We select data based on quality ratings to train language models from scratch, and discover the importance of balancing quality and diversity in data selection. Using quality ratings as selection criteria, our models outperform models trained on randomly selected data/data selected using other methods on perplexity and in-context learning task performance. Further analysis delves into the ratings' characteristics, biases, and broader implications, including how they suppress/promote texts associated with certain languages, regions, topics, or social roles. The talk concludes with a study of the offensiveness of selected texts using toxicity models.

1 0

Zheng Shi will present her FPO "Task-Specific Computational Cameras" on Friday, May 3, 2024 at 3:00 PM in CS 302.
by CS Grad Department 26 Apr '24

26 Apr '24

Zheng Shi will present her FPO "Task-Specific Computational Cameras" on Friday, May 3, 2024 at 3:00 PM in CS 302. Location: CS 302 The members of Zheng’s committee are as follows: Examiners: Felix Heide (Adviser), Adam Finkelstein, Olga Russakovsky Readers: Ellen Zhong, Tian-Ming Fu A copy of his thesis is available upon request. Please email gradinfo(a)cs.princeton.edu if you would like a copy of the thesis. Everyone is invited to attend her talk. Abstract follows below: Machine vision, while fundamentally relying on images as inputs, has traditionally treated image acquisition and image processing as two separate tasks. However, traditional image acquisition systems are tuned for aesthetics, photos that please the human eye, and not computational tasks that requires beyond human vision. My research focuses on developing task-specific computational imaging systems to enable the capture of information that extends beyond the capabilities of standard RGB cameras, thereby enhancing the effectiveness of downstream machine vision applications. This thesis begins with combining multiple imaging modalities to facilitate training on unpaired real-world datasets, addressing the scarcity of supervised training data. We introduce ZeroScatter, a single-image descattering method capable of removing adverse weather effects from RGB captures. By integrating model-based, temporal, and multi-view cues, as well as information contained in gated imager captures, we offer indirect supervision for training on real-world adverse weather captures lacking ground truth. This approach significantly enhances generalizability on unseen data, surpassing methods trained exclusively on synthetic adverse weather data. Despite its great applicability, relying solely on conventional RGB image inputs limits available information, and requires the model to fill in gaps by generating plausible inferences based on learnt prior, such as when car window wiper obscure objects from the dash cameras. To bypass these constraints, we shift towards computational cameras, and design specialized flat optics to boost the capabilities of cameras for a range of applications. We first propose a computational monocular camera that optically cloaks unwanted near-camera obstructions. We learn a custom diffractive optical element (DOE) that performs depth-dependent optical encoding, scattering nearby occlusions while allowing paraxial wavefronts emanating from background objects to be focused. This allows us to computationally reconstruct unobstructed images without requiring captures different camera views or hallucinations. Lastly, we introduce a split-aperture 2-in-1 computational camera that combines application-specific optical modulation with conventional imaging into one system. This approach simplifies complex inverse problems faced by computational cameras, enhances reconstruction quality, and offers a real-time viewfinder experience; paving the way for the adoption of computational camera technology in consumer devices.

1 0