- talks - lists.cs.princeton.edu

Wuwei Zhang will present her MSE Talk"Improving Long-Context Reasoning with Query-Focused Retrieval Heads" onTuesday, April 21, 2026 at 4:00p in CS 302.
by CS Grad Department 16 Apr '26

16 Apr '26

Wuwei Zhang will present her MSE Talk "Improving Long-Context Reasoning with Query-Focused Retrieval Heads" on Tuesday, April 21, 2026 at 4:00p in CS 302. Committee Members: Prof. Danqi Chen (advisor),Prof. Karthik Narasimhan (reader). All are welcome to attend. Title: Improving Long-Context Reasoning with Query-Focused Retrieval Heads Abstract: Large language models (LLMs) have demonstrated strong performance across many tasks, yet they often struggle to effectively utilize information from long contexts. Recent studies suggest that a subset of attention heads, known as retrieval heads, plays a critical role in retrieving relevant information from long contexts. However, existing methods for identifying these heads rely on copy-paste behavior in synthetic tasks, which may not reflect real-world retrieval requirements. In this thesis, we investigate how retrieval heads contribute to long-context reasoning and propose methods to better utilize them during inference. First, we introduce Query-Focused Retrieval Heads (QRHead), a method for identifying attention heads that selectively retrieve information relevant to the input query. By aggregating the accumulated attention mass of QRHead, we further develop QRRetriever, a training-free retrieval approach that selects the most relevant context segments for downstream reasoning tasks. Building on QRHead, we propose Dynamic Attention-Scaling Decoding (DySCO), an inference-time method that dynamically rescales attention based on query-focused retrieval signals, enabling models to more effectively access relevant information during generation. Our methods achieve substantial improvements on long-context reasoning benchmarks across multiple model families, without requiring additional training. Together, these results demonstrate that retrieval heads provide a useful lens for understanding and improving long-context reasoning in LLMs, and that inference-time interventions guided by these mechanisms can lead to more effective context utilization.

1 0

Amelia Dobis will present her General Exam "New Paradigms for Hardware Design, Compilation, and Verification" on April 24, 2026 at 1pm at 194 Nassau St, 2nd floor meeting area (Suite 21) and Zoom.
by Gradinfo 15 Apr '26

15 Apr '26

Amelia Dobis will present her General Exam "New Paradigms for Hardware Design, Compilation, and Verification" on April 24, 2026 at 1pm at 194 Nassau St, 2nd floor meeting area (Suite 21) and Zoom. Committee: Mae Milano (adviser), Aarti Gupta, Sneha Goenka Zoom: [ https://princeton.zoom.us/j/99419002211?pwd=hd5f2sSwZUt7yfU3bBTNOILyAKz6tE.1 | https://princeton.zoom.us/j/99419002211?pwd=hd5f2sSwZUt7yfU3bBTNOILyAKz6tE.1 ] Everyone is invited to attend the talk, and those faculty wishing to remain for the oral exam following are welcome to do so. [ https://docs.google.com/document/d/1TdYi6TAy0gCCj46ryReP9Ms9C-ila6dQ8dNhuCF… | Abstract & Reading List: https://docs.google.com/document/d/1TdYi6TAy0gCCj46ryReP9Ms9C-ila6dQ8dNhuCF… ]

1 0

Sofiia Druchyna will present her MSE Talk "Spectral Credit Reconstruction for GRPO in Long-Horizon RLHF" on Monday, April 20, 2026 at 10:30a in CS 401.
by CS Grad Department 15 Apr '26

15 Apr '26

Sofiia Druchyna will present her MSE Talk "Spectral Credit Reconstruction for GRPO in Long-Horizon RLHF" on Monday, April 20, 2026 at 10:30a in CS 401. Committee Members: Prof. Elad Hazan (advisor), Prof. Manoel Horta Ribeiro (reader). All are welcome to attend. Title: Spectral Credit Reconstruction for GRPO in Long-Horizon RLHF Abstract: Reinforcement Learning from Human Feedback (RLHF) relies on sequence-level rewards to train large language models, yet policy updates are typically applied uniformly across all tokens in a generated completion. This uniform credit assignment ignores the temporal structure of language generation and contributes to high-variance gradient estimates in long-horizon settings. We propose a spectral credit assignment method for Group-Relative Policy Optimization (GRPO) that redistributes scalar advantages across token positions using a low-dimensional spectral parameterization. Our approach represents token-level credit as a smooth combination of fixed low-frequency basis functions, with mixture weights predicted by a state-dependent gating network operating on pooled hidden states. This construction preserves the expected GRPO update while constraining token-level variation to a controlled, low-frequency subspace, thereby reducing variance without introducing additional critics or token-level rewards. We show that our method interpolates smoothly between standard GRPO and structured token-level credit assignment, offering a simple and stable mechanism for incorporating temporal structure into RLHF training.

1 0

Phyllis Wang will present her MSE talk "Implementation and Scaling of Quantum Cognitive Models" on Thursday, April 23rd, 2026 at 1pm in CS 301.
by Gradinfo 15 Apr '26

15 Apr '26

Phyllis Wang will present her MSE talk "Implementation and Scaling of Quantum Cognitive Models" on Thursday, April 23rd, 2026 at 1pm in CS 301. Adviser: Margaret Martonosi, Reader: Raghav Pothukuchi All are welcome to attend Title: Implementation and Scaling of Quantum Cognitive Models Abstract: The computational aspect of cognitive models inspired by quantum mechanics poses unique considerations in their implementation, both through classical and quantum computing. We examine the quantum well-inspired Multi-Particle Multi-Well (MPMW) model for decision making to find how quantum cognitive techniques scale on various types of hardware and modeling approaches. This thesis explores an end-to-end implementation of the MPMW model, including the representation of quantum wells as the landscape becomes more discretized, addressing the challenge of memory bottlenecks that arise from representing quantum states. We further explore the computational and memory scaling limits of finding the eigenstates of the wells, and analyze the results of scaling on the model predictions. Finally, we look at the role of quantum computing and the possibility of scaling model sizes further through the use of the hybrid quantum-classical Variational Quantum Eigensolver algorithm.

1 0

Alan Zhang will present his MSE talk "Detection-based Attention Steering System for Vision Language Models" on Friday, Apr 24th, 2026 in CS 105 at 2pm in CS 105.
by Gradinfo 15 Apr '26

15 Apr '26

Alan Zhang will present his MSE talk "Detection-based Attention Steering System for Vision Language Models" on Friday, Apr 24th, 2026 in CS 105 at 2pm in CS 105. All are welcome to attend. Presenter: Alan Zhang (az0686) Advisor: Ravi Netravali Reader: Wyatt Lloyd Abstract: Modern Vision Language Models (VLMs) have demonstrated remarkable versatility in a wide variety of applications involving multimodal reasoning, captioning, and analysis. However, VLMs still struggle with the fundamental tasks of object detection and localization, especially when compared to traditional Convolutional Neural Network (CNN) models. Contemporary VLMs tend to allocate very scarce attention to non-textual inputs such as image tokens and lack the visual inductive biases present in CNNs, making it difficult for them to identify challenging objects in complex scenes. To address these limitations, we propose a detection-based dynamic attention steering system that utilizes the locality insight from convolutions to efficiently steer a VLM's attention toward more relevant sections of an image outlined by bounding boxes from a CNN-based detector model. The steering intensity during each inference is dynamically scaled according to the confidence of the bounding boxes. Extensive evaluations across multiple state-of-the-art VLMs demonstrate substantial improvements on object localization-oriented benchmarks, achieving up to a 4% accuracy gain. Results show the effectiveness of combining different model architectures to harness their respective strengths for advancing VLM capabilities.

1 0

Ziyu Xiong will present her General Exam "Atomic Diffusion Models for Small Molecule Structure Elucidation from NMR Spectra" on Friday, April 17, 2026 at 3:00 PM in Friend 202 and via zoom.
by CS Grad Department 14 Apr '26

14 Apr '26

Ziyu Xiong will present her General Exam "Atomic Diffusion Models for Small Molecule Structure Elucidation from NMR Spectra" on Friday, April 17, 2026 at 3:00 PM in Friend 202 and via zoom. Zoom link: https://princeton.zoom.us/j/95476730233 Committee Members: Ellen Zhong (advisor), Zhuang Liu, Ryan Adams Abstract: Nuclear Magnetic Resonance (NMR) spectroscopy is a cornerstone technique for determining the structures of small molecules and is especially critical in the discovery of novel natural products and clinical therapeutics. Yet, interpreting NMR spectra remains a time-consuming, manual process requiring extensive domain expertise. We introduce ChefNMR (CHemical Elucidation From NMR), an end-to-end framework that directly predicts an unknown molecule's structure solely from its 1D NMR spectra and chemical formula. We frame structure elucidation as conditional generation from an atomic diffusion model built on a non-equivariant transformer architecture. To model the complex chemical groups found in natural products, we curated a dataset of simulated 1D NMR spectra for over 111,000 natural products. ChefNMR predicts the structures of challenging natural product compounds with an unsurpassed accuracy of over 65% on simulated benchmarks, but generalizing to real experimental NMR spectra remains a key challenge. To bridge this synthetic-to-real gap, we propose further improvements, including a more robust peak-based spectral encoder, a flow matching generative framework, a multimodal transformer architecture, and expanded training data. This work takes a significant step toward solving the grand challenge of automating small-molecule structure elucidation and highlights the potential of deep learning in accelerating molecular discovery. Reading List: https://docs.google.com/document/d/1OLM7gak0-XaH8To0Al_zOrEIwyvezDM4bAxvUyO… 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

Jens Tuyls will present his FPO "From Expert Imitation to Autonomous Discovery in Intelligent Agents" on Tuesday, April 21, 2026 at 2 PM in CS 105.
by CS Grad Department 14 Apr '26

14 Apr '26

Jens Tuyls will present his FPO "From Expert Imitation to Autonomous Discovery in Intelligent Agents" on Tuesday, April 21, 2026 at 2 PM in CS 105. The members of Jens’ committee are as follows: Examiners: Karthik Narasimhan (Adviser), Benjamin Eysenbach, Tom Silver Readers: Danqi Chen, Chi Jin 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: Learning from demonstrations and learning from experience are two learning paradigms that lie at the heart of today’s most capable AI systems. This thesis focuses on questions related to both of these paradigms: (1) how to recover expert behavior from demonstration data, (2) how to explore an environment without access to demonstration data, and (3) how to leverage expert knowledge to guide exploration. Chapter 1 focuses on learning from expert demonstrations and asks the following question: What role do model and data size play in recovering the expert’s performance when training policies with behavioral cloning? We demonstrate that the same empirical scaling laws used in language modeling accurately describe the loss and return functions in behavioral cloning on single-agent games, allowing practitioners to predict the compute-optimal model and data size to reach a certain level of performance. Chapters 2 and 3 introduce two novel reinforcement learning-based methods for exploring an environment without access to expert demonstrations. The first method uses explicitly disentangled exploration and exploitation phases combined with curiosity-based intrinsic rewards to explore the state space, setting a new state-of-the-art in challenging text game environments. The second method learns a large set of policies (or “skills”) through mutual information skill learning that, taken together, cover the state space of the MDP. In continuous control tasks, the method is competitive with current state-of-the-art skill-learning methods. Finally, Chapter 4 studies the interplay between learning from expert data and autonomous exploration by asking the following question: Can the knowledge in language model representations, which is mostly obtained from expert data, help drive exploration toward learning novel behaviors? Our empirical results suggest deliberate exploration with reinforcement learning on top of pre-trained language models can be a promising path for learning novel behaviors beyond those of human experts. Taken together, this thesis provides foundational empirical insight for learning from expert data as well as novel empirical methods for learning from experience through exploration.

1 0

Jane Castleman will present her MSE thesis "Measuring Validity in Resume Screening Systems" on Tuesday, April 14th, 2026 in Sherrerd Hall 306 at 4:30pm.
by Gradinfo 13 Apr '26

13 Apr '26

Jane Castleman will present her MSE thesis "Measuring Validity in Resume Screening Systems" on Tuesday, April 14th, 2026 in Sherrerd Hall 306 at 4:30pm. Advisor: Aleksandra Korolova Reader: Lydia T. Liu All are welcome to attend. Abstract: Resume screening is perceived as a particularly suitable task for LLMs given their ability to analyze natural language; thus many entities rely on general purpose LLMs without further adapting them to the task. While researchers have shown that some LLMs are biased in their selection rates of different demographics, studies measuring the validity of LLM decisions are limited. One of the difficulties in externally measuring validity stems from lack of access to a large corpus of resumes for whom the ground truth in their ranking is known and that has not already been used for LLM training. In this work, we overcome this challenge by systematically constructing a large dataset of resumes tailored to particular jobs that are directly comparable, with a known ground truth of superiority. We then use the constructed dataset to measure the validity of ranking decisions made by various LLMs, finding that many models are unable to consistently select the resumes describing more qualified candidates. Furthermore, when measuring the validity of decisions, we find that models do not reliably abstain when ranking equally-qualified candidates, and select candidates from different demographic groups at different rates. Our proposed framework provides a principled approach to audit LLM resume screeners in the absence of ground truth, offering a crucial tool to independent auditors and developers to ensure the validity of these systems as they are deployed.

1 0

CoE Seminar - Prof. David Sholl, 4/23 at 10:00 am in Maeder Hall
by Emily C. Lawrence 09 Apr '26

09 Apr '26

CoE Seminar Thursday, April 23 at 10:00am in Maeder Hall Auditorium David Sholl, Rice University Title: When Do I Get My Flying Car? Assessing the Potential of AI to Revolutionize Separations Science Abstract: Artificial intelligence and machine learning have been suggested as tools that will revolutionize many areas of science and society. I will discuss the potential and pitfalls for these methods in a key field within chemical engineering, namely separations science. Examples will include scanning collections of thousands of porous materials for adsorption-based separations of complex molecular mixtures, searching for high performance materials for Direct Air Capture of carbon dioxide, and developing mixed matrix membranes for gas separations.

1 0

Peter Zoller presents a public lecture on quantum computing on April 14 at 5 pm
by Emily C. Lawrence 09 Apr '26

09 Apr '26

PCTS Public Lecture Tuesday, April 14, 2026 5:00 PM, McDonnell Hall, Room A-02 Speaker: Professor Peter Zoller Title: When Atoms Compute Subtitle: Quantum computing, simulation, and sensing with atoms, ions, and light Abstract: Quantum mechanics governs nature at its smallest scales—but today we can engineer that quantum world by creating and controlling entanglement, the unique fingerprint of quantum physics. Using laser-controlled atoms and ions, researchers can build programmable quantum devices that compute beyond classical limits, emulate complex quantum matter, and approach fundamental quantum limits. This talk gives a non-technical overview of leading platforms, from trapped-ion processors to quantum simulators based on ultracold atoms. We illustrate the ideas with snapshots from current research: string-breaking dynamics inspired by high-energy physics, using simulators to guide materials design, methods to validate what quantum devices are doing when classical calculation is no longer feasible, and an outlook toward programmable quantum sensors. There is no registration requirement for the public lecture.

1 0