- talks - lists.cs.princeton.edu

Ameet Deshpande will present his FPO "From Insights to Improvements: Advancements across the LLM Lifecycle" on Wednesday, November 12, 2025 at 1:30 PM in Louis A. Simpson International Building - A71.
by CS Grad Department 07 Nov '25

07 Nov '25

Ameet Deshpande will present his FPO " From Insights to Improvements: Advancements across the LLM Lifecycle " on Wednesday, November 12, 2025 at 1:30 PM in Louis A. Simpson International Building - A71 . The members of Ameet ’s committee are as follows: Examiners: Karthik Narasimhan (Adviser), Danqi Chen , Sanjeev Arora Readers: Andrés Monroy-Hernández , Suma Bhat 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 his talk. Abstract follows below: Large language models (LLMs) have transformed natural language understanding and generation, yet their development remains a fragmented process. Each stage in the LLM lifecycle—pre-training, post-training, evaluation, and improvement—poses distinct challenges that are rarely studied together. This thesis pushes the envelope on all phases while also tying them together for continuous adaptation of models. We begin with the pre-training phase, analyzing what enables cross-lingual generalization in multilingual models, where pre-training on one language bestows skills in another. In the posttraining phase, we show that instruction-tuned LLMs exhibit biases related to the persona they are simulating, highlighting the need for socially aware alignment methods. Turning to evaluation, we demonstrate that existing semantic similarity benchmarks fail to capture contextual reasoning. We propose Conditional STS, a framework that measures similarity relative to explicit conditions, producing fine-grained feedback. Finally, we introduce QUALEVAL, a system that closes the loop between evaluation and improvement by automatically surfacing qualitative insights and guiding targeted model refinement. Together, these contributions reimagine the LLM lifecycle as a continuous feedback process—one where insights from each stage inform the next. By integrating pre-training, alignment, evaluation, and improvement, this thesis moves toward the vision of adaptive, self-improving language models.

1 0

CS Colloquium Speaker: Alexandra Ion, Monday, Oct 27 at 12:10pm
by Emily C. Lawrence 07 Nov '25

07 Nov '25

CS Colloquium Speaker Speaker: Alexandra Ion, Carnegie Mellon University Date: Monday, October 27 Time: 12:10pm EST Location: CS 105 Host: Parastoo Abtahi Event page: [ https://www.cs.princeton.edu/events/human-centered-physical-ai | https://www.cs.princeton.edu/events/human-centered-physical-ai ] Register for live-stream online here: [ https://princeton.zoom.us/webinar/register/WN_q7rNj3QZQ7q3Uf5aSYB4Dw | https://princeton.zoom.us/webinar/register/WN_q7rNj3QZQ7q3Uf5aSYB4Dw ] Title: Human-centered Physical AI Abstract: The world around us is inherently physical, yet adaptive interfaces focus mostly on digital content and representations. Physical AI is moving in the right direction by creating models that can understand instructions and perform physical tasks in the real world, typically with humanoid or quadrupedal robots as physical AI agents. In this talk, I envision a future where such physical AI agents move into the background—instead of interacting with large general-purpose robots, we should interact with physical objects that are familiar to us. I call this *Unobtrusive Physical AI*. Bottles, desks, chairs, walls; any object surrounding us should dynamically adapt to our needs, by not only adjusting their user interfaces to fit the context but also by transforming their physical features, material properties, and affordances to instantly become exactly what users need in that moment. I will discuss how creating such intelligent agents requires two main components: (1) adaptive physical architectures to facilitate physical change, e.g., through metamaterials, shape-changing interfaces, soft robotics, etc., and (2) sensing and prediction systems to understand when and what functionality users require. With new approaches in predictive user modeling and innovations in manufacturing and material science, the time might just be ripe to make this challenging vision a reality. Bio: I am an Assistant Professor at the Human-Computer Interaction Institute at Carnegie Mellon University's School of Computer Science. I direct the Interactive Structures Lab, where we investigate and develop physical interfaces that adapt to users' needs. My lab’s work is published at and awarded by top-tier HCI (ACM CHI & UIST) and graphics venues (ACM SIGGRAPH) and has been recognized with the prestigious NSF CAREER award. It was showcased in multiple exhibitions, including a permanent exhibition at the Ars Electronica Center in Austria. Our work also captured the interest of media such as Wired, Dezeen, Fast Company, Gizmodo, etc., and presented at TEDx. Before joining CMU, I was a postdoctoral researcher at ETH Zurich in Computer Graphics and completed my PhD in HCI at the Hasso Plattner Institute, a small, highly selective, top-tier institute for computer science in Germany.

1 2

Reminder: PPH seminar: Digital Phenotyping for Precision Mental Health - tomorrow at 12pm
by Emily C. Lawrence 06 Nov '25

06 Nov '25

Please join the Princeton Precision Health (PPH) Initiative for the next talk of our Fall 2025 semester series on Friday, November 7th, at 12:00 pm, at 252 Nassau Street . Dr. Nelson Freimer is Director of UCLA’s Depression Grand Challenge, a campus-wide initiative that aims to cut the burden of depression in half by 2050. He will present a seminar on, Digital Phenotyping for Precision Mental Health. Widespread application of digital phenotyping – assessments using data passively obtained from smartphones and wearables – could transform mental health research and clinical care by providing nearly continuous, quantitative, and objective information across broad physiological, behavioral, and emotional domains. The UCLA-Apple Digital Mental Health Study (DMHS) is the largest and longest digital mental health study conducted to date, having collected up to12 months of sensor data from phones and watches (resulting in >900 phenotypic features) in a sample of >4,000 participants, diverse by age, sex at birth, ethnicity, and depression symptom severity. Analyses of DMHS data identify highly specific associations between depression and anxiety symptoms and sensor domains. Additionally, the results suggest that longitudinal digital phenotyping will enable dissection of heterogenous diagnostic categories, a key step towards precision mental health. Lunch will be provided. Getting to the seminar space currently requires that you climb a set of stairs. If an accommodation is needed, please contact PPH in advance at: [ mailto:PrincetonPPH@princeton.edu | PrincetonPPH(a)princeton.edu ]

1 0

TOMORROW: AI Lab Policy Fellows Seminar 2:00 PM, Lewis Library 120
by Emily C. Lawrence 05 Nov '25

05 Nov '25

[ https://citp.princeton.edu/events/2025/ai-lab-policy-fellows-seminar | https://citp.princeton.edu/events/2025/ai-lab-policy-fellows-seminar ] AI Lab Policy Fellows Seminar Thursday, November 6, 2025 2 – 4 p.m. Lewis Library 120 Open to Princeton University Faculty, staff, and students The AI Lab is hosting the Policy Fellows Seminar with speakers Aisha Nájera, Neil Perry, Ryan Steed, and Rajesh Veeraraghavan. As part of the AI Lab’s Policy Fellows Program, these researchers work closely with governmental agencies at the state and federal level to advance the responsible development and utilization of AI. Join us as they share insights from their work at the intersection of technology and public policy. AI Lab Policy Fellows: Aisha Nájera and Rajesh Veeraraghavan Talk Title: Coding for Safety and Expertise: Assessing Government AI in Health and Human Services Abstract: We analyze a database of government AI use cases in welfare to ask what kinds of risks and forms of expertise are made visible, and which remain opaque. By coding these cases against emerging safety categories, we sketch a mapping schema that links AI functions to potential risks while also tracing how expertise is implicated, redistributed, or obscured in these systems. In doing so, we also consider how existing AI risks taxonomies can be assessed and adapted to the specific operational needs, identifying practices for managing risks in context. Our aim is to open a way of thinking about how welfare AI inventories surface questions of safety and expertise. Neil Perry Talk Title: A Measurement Platform for Prompt Injection in AI Agents Abstract: My research at CAISI focuses on measuring the security of AI agents against prompt injection to better understand current vulnerabilities and defenses. I am designing a crowdsourced platform that captures the creative breadth of attacks and defense designs from real people, provides a continuously updated leaderboard, and enables large-scale comparative studies. The platform makes it easy for anyone to write and share modular agent architectures, which can then be evaluated for formal security properties. By making measurement an ongoing, community-driven effort, the system captures new attacks and defenses as they emerge and compares the performance of different models and architectures to better understand how to use and secure agents. Ryan Steed Talk Title: Does It Work? Open Measurement Science Questions for AI Evaluation Abstract: Integrating AI systems into real-world workflows is a complex process that can have unexpected effects. Industry-standard benchmarks fail to capture these complexities—they generally involve simplistic tasks in isolated settings. If our most popular measures may not tell us much about performance in the real world, how should we measure whether AI systems really work? How do we evaluate risks associated with their use? In this talk, I’ll share some open measurement science questions facing policymakers and talk about the future of evaluation and procurement standards in the workplace. Sponsorship of an event does not constitute institutional endorsement of external speakers or views presented.

1 0

Vishvak Murahari will present his FPO "Towards efficient and personalized Generative AI" on Wednesday, November 12, 2025 at 11:00 AM in Friend 009.
by CS Grad Department 05 Nov '25

05 Nov '25

Vishvak Murahari will present his FPO "Towards efficient and personalized Generative AI" on Wednesday, November 12, 2025 at 11:00 AM in Friend 009. The members of Vishvak’s committee are as follows: Examiners: Karthik Narasimhan (Adviser), Danqi Chen, Benjamin Eysenbach Readers: Kai Li, Elad Hazan 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 his talk. Abstract follows below: The rapid evolution of large language models (LLMs) has redefined generative artificial intelligence. Yet, modern systems remain constrained by three persistent challenges: their immense computational cost, their opaque evaluation, and their limited personalization. This dissertation addresses these challenges through a unifying goal—to make generative AI both efficient and personalized—advancing methods that improve computational efficiency, interpretability, and human alignment. The first part of this work introduces architectural techniques that unlock latent capacity within neural networks. Data Multiplexing (DataMUX) enables models to process multiple inputs concurrently, improving inference throughput without sacrificing accuracy. Building on this idea, MUX-PLMs extend multiplexing to pretrained language models, demonstrating scalable multi-input multi-output (MIMO) inference across standard NLP benchmarks. The second part of the dissertation shifts from computation to comprehension. QualEval introduces a qualitative evaluation framework that replaces single-number metrics with interpretable, domain-level insights. By automatically discovering latent task domains and generating diagnostic reports, QualEval bridges the gap between benchmarking and guided improvement, enabling models that are transparent and improvable. The third part examines how generative systems shape modern information ecosystems. Generative Engine Optimization (GEO) formalizes visibility in AI search as a black-box optimization problem, introducing impression metrics and a benchmark, GEO-Bench. PersonaGym then provides a large-scale framework for evaluating persona-conditioned agents through normative, prescriptive, and descriptive dimensions. Across hundreds of simulated personas, it quantifies personality coherence and bias, establishing a foundation for context-aware, human-aligned generative models. Together, these contributions trace a unified trajectory toward efficient personalization—a paradigm in which generative systems learn economically, explain their reasoning clearly, and engage with humans authentically.

1 0

Special Quantum Seminar: Eli Goldin (NYU): "One-Way Puzzles: A Central Primitive for Quantum Cryptography" 2 pm, Mon. Nov. 10 (Comp Sci, 105)
by Emily C. Lawrence 05 Nov '25

05 Nov '25

Quantum Seminar Speaker: Eli Goldin, NYU Date: Monday, Nov. 10 Time: 2:00 pm Location: Computer Science Building, CS105 Host: Prof. Alex Lombardi, CS Title: One-Way Puzzles: A Central Primitive for Quantum Cryptography" Abstract: One-way functions are generally seen as a fundamental (or "central") primitive for classical cryptography. The existence of nearly any cryptography implies the existence of one-way functions, and one-way functions satisfy a number of useful properties which make them easy to work with. In the quantum setting, we give evidence that a recently introduced primitive, the one-way puzzle, is a natural analogue of one-way functions in the quantum regime. In particular, they are a useful tool for understanding the fundamental hardness behind quantum cryptography. Towards this end, we show that one-way puzzles satisfy nearly all the "nice properties" of one-way functions. We present a one-way puzzle combiner, a universal construction, a security amplification theorem, and show that one-way puzzles are equivalent to many natural variants. As a result of these nice properties, we demonstrate that one-way puzzles can be constructed from a large class of quantum cryptographic primitives (in particular, those broken by a #P oracle). Furthermore, we show a number of interesting connections between one-way puzzles and the fields of meta-complexity and (efficiently verifiable) quantum advantage, and discuss black-box separations between one-way puzzles and other primitives.

1 0

Dan Friedman will present his FPO "Algorithmic Interpretability for Language Models" on Tuesday, November 11, 2025 at 1:00 PM in AI Lab Room 274 and Zoom.
by CS Grad Department 04 Nov '25

04 Nov '25

Dan Friedman will present his FPO " Algorithmic Interpretability for Language Models " on Tuesday, November 11, 2025 at 1:00 PM in AI Lab Room 274 and Zoom. Location: Zoom link: https://www.google.com/url?q=https%3A%2F%2Fprinceton.zoom.us%2Fj%2F86367410… The members of Dan ’s committee are as follows: Examiners: Danqi Chen (Adviser), Tom Griffiths , Adji Bousso Dieng Readers: Sanjeev Arora , Alexander Rush (Cornell University) 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 his talk. Abstract follows below: Modern natural language processing systems are black boxes. As a result, these models are fundamentally difficult to audit, debug, or trust. Standard post-hoc interpretability methods offer partial insight into model behavior but do not provide complete or faithful descriptions of the underlying computation. The aim of my research is to achieve an “algorithm-level” understanding of language processing systems, by aligning neural models with human-readable, rule-based programs. This thesis is organized into two parts. The first part involves developing intrinsically interpretable methods—machine learning models that are designed from the outset to be easily converted into symbolic form. I will present a method for finding interpretable features in text data sets using probabilistic grammars, which can recover “shortcuts” in NLP classification datasets—shallow patterns in the training data that are correlated with the target labels but do not generalize reliably. Next, I will introduce my work on developing transformers that are mechanistically interpretable by design, which can be trained using gradient-based optimization and then automatically converted into discrete, human-readable programs. These Transformer Programs are capable of learning effective solutions for a variety of tasks, and can be analyzed using standard code-analysis tools. The second part of the thesis involves aligning existing black-box models with symbolic rules, either by theoretical analysis or by extracting rule-based descriptions of their behavior. This includes a theoretical and empirical analysis of how transformers can represent classic, rule-based chatbot algorithms like ELIZA, and a method for extracting rule-based explanations of attention features using Sparse Autoencoders. I will conclude by outlining a roadmap for future research towards the goal of algorithm-level understanding of neural language models.

1 0

PPH seminar: Digital Phenotyping for Precision Mental Health - Friday, Nov 7th at 12pm
by Emily C. Lawrence 03 Nov '25

03 Nov '25

Please join the Princeton Precision Health (PPH) Initiative for the next talk of our Fall 2025 semester series on Friday, November 7th, at 12:00 pm, at 252 Nassau Street. Dr. Nelson Freimer is Director of UCLA’s Depression Grand Challenge, a campus-wide initiative that aims to cut the burden of depression in half by 2050. He will present a seminar on, Digital Phenotyping for Precision Mental Health. Widespread application of digital phenotyping – assessments using data passively obtained from smartphones and wearables – could transform mental health research and clinical care by providing nearly continuous, quantitative, and objective information across broad physiological, behavioral, and emotional domains. The UCLA-Apple Digital Mental Health Study (DMHS) is the largest and longest digital mental health study conducted to date, having collected up to12 months of sensor data from phones and watches (resulting in >900 phenotypic features) in a sample of >4,000 participants, diverse by age, sex at birth, ethnicity, and depression symptom severity. Analyses of DMHS data identify highly specific associations between depression and anxiety symptoms and sensor domains. Additionally, the results suggest that longitudinal digital phenotyping will enable dissection of heterogenous diagnostic categories, a key step towards precision mental health. Lunch will be provided. Getting to the seminar space currently requires that you climb a set of stairs. If an accommodation is needed, please contact PPH in advance at: [ mailto:PrincetonPPH@princeton.edu | PrincetonPPH(a)princeton.edu ]

1 0

Blockchain and Belief: Histories and Futures of Money, Power, and Faith
by Emily C. Lawrence 29 Oct '25

29 Oct '25

Please join the Princeton University DeCenter and the Princeton Project in Philosophy & Religion for a public panel discussion exploring the intersections of blockchain technology, religion, and power. This conversation will examine how historical and religious frameworks can shed light on emerging blockchain phenomena and offer ethical guidance in times of shifting authority and trust. The event will take place on November 5th at 4:30 p.m. in the Friend Center Convocation Room (113) . Panelists include Devin Singh (Dartmouth College), David Miller (Princeton University), Gordon Gant (Chi-Squared), and Meltem Demirors (Crucible Capital). The discussion will be moderated by Andrew Chignell (Princeton University). This event is free and open to the public. If you are interested in attending the full-day workshop, please email [ mailto:decenter@princeton.edu | decenter(a)princeton.edu ] . Best, Madelyn LaBuda Administrative and Events Coordinator DeCenter Engineering Quad, C215 Princeton University E: ml2512 [ mailto:email@princeton.edu | @princeton.edu ] M: 570-599-5338

1 0

CS Colloquium speaker: Alexandra Ion TODAY at 12:10pm
by Emily C. Lawrence 27 Oct '25

27 Oct '25

CS Colloquium Speaker Speaker: Alexandra Ion, Carnegie Mellon University Date: Monday, October 27 Time: 12:10pm EST Location: CS 105 Host: Parastoo Abtahi Event page: https://www.cs.princeton.edu/events/human-centered-physical-ai Register for live-stream online here: https://princeton.zoom.us/webinar/register/WN_q7rNj3QZQ7q3Uf5aSYB4Dw Title: Human-centered Physical AI Abstract: The world around us is inherently physical, yet adaptive interfaces focus mostly on digital content and representations. Physical AI is moving in the right direction by creating models that can understand instructions and perform physical tasks in the real world, typically with humanoid or quadrupedal robots as physical AI agents. In this talk, I envision a future where such physical AI agents move into the background—instead of interacting with large general-purpose robots, we should interact with physical objects that are familiar to us. I call this *Unobtrusive Physical AI*. Bottles, desks, chairs, walls; any object surrounding us should dynamically adapt to our needs, by not only adjusting their user interfaces to fit the context but also by transforming their physical features, material properties, and affordances to instantly become exactly what users need in that moment. I will discuss how creating such intelligent agents requires two main components: (1) adaptive physical architectures to facilitate physical change, e.g., through metamaterials, shape-changing interfaces, soft robotics, etc., and (2) sensing and prediction systems to understand when and what functionality users require. With new approaches in predictive user modeling and innovations in manufacturing and material science, the time might just be ripe to make this challenging vision a reality. Bio: I am an Assistant Professor at the Human-Computer Interaction Institute at Carnegie Mellon University's School of Computer Science. I direct the Interactive Structures Lab, where we investigate and develop physical interfaces that adapt to users' needs. My lab’s work is published at and awarded by top-tier HCI (ACM CHI & UIST) and graphics venues (ACM SIGGRAPH) and has been recognized with the prestigious NSF CAREER award. It was showcased in multiple exhibitions, including a permanent exhibition at the Ars Electronica Center in Austria. Our work also captured the interest of media such as Wired, Dezeen, Fast Company, Gizmodo, etc., and presented at TEDx. Before joining CMU, I was a postdoctoral researcher at ETH Zurich in Computer Graphics and completed my PhD in HCI at the Hasso Plattner Institute, a small, highly selective, top-tier institute for computer science in Germany.

1 0