Weekly Media Update

Week of December 12, 2025

Warren Center Faculty Affiliates at NeurIPS 2025

Fifty-five papers from Warren Center faculty affiliates, including four Spotlight papers, have been accepted to the Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS 2025). A huge congratulations to these affiliates for their impressive research. See more about the papers in the links below.

Learning From Frustration: Torsor CNNs on Graphs
12/7: NeurIPS 2025 Workshop Paper | Robert Ghrist

T-FIX: Text-Based Explanations With Features Interpretable to eXperts
12/7: NeurIPS 2025 Workshop Paper | Bhuvnesh Jain and Lyle Ungar

The Binding Problem in Vision Models: Geometric, Functional, and Behavioral Approaches
12/7: NeurIPS 2025 Workshop Paper | Konrad Kording

A Constrained Optimization Perspective of Unrolled Transformers
12/7: NeurIPS 2025 Workshop Paper | Alejandro Ribeiro

Augmented Lagrangian for Constrained Learning
12/7: NeurIPS 2025 Workshop Paper | Alejandro Ribeiro

Infinity Embeddings: Representation Learning with Ultrametric Structure for Faster Search via Constrained Learning
12/7: NeurIPS 2025 Workshop Talk | Alejandro Ribeiro

Transferability of Graph Transformers With Convolutional Positional Encodings
12/7: NeurIPS 2025 Workshop Paper | Alejandro Ribeiro

Unrolled Neural Networks for Constrained Optimization
12/7: NeurIPS 2025 Workshop Paper | Alejandro Ribeiro

Online Omniprediction With Long-Term Constraints
12/7: NeurIPS 2025 Workshop Paper | Aaron Roth

GNN-Parametrized Diffusion Policies for Wireless Resource Allocation
12/7: NeurIPS 2025 Workshop Paper | Shirin Saeedi Bidokhti and Alejandro Ribeiro

Measuring Dependencies Between Biological Signals With Temporal Self-Supervision, and Its Limitations
12/7: NeurIPS 2025 Workshop Paper | Ted Satterthwaite

MARSHA: Multi-Agent RAG System for Hazard Adaptation
12/7: NeurIPS 2025 Workshop Paper | Weijie Su

Culturally-Aware Conversations: A Framework & Benchmark for LLMs
12/7: NeurIPS 2025 Workshop Paper | Lyle Ungar

Know Me, Respond to Me: Benchmarking LLMs for Dynamic User Profiling and Personalized Responses at Scale
12/7: NeurIPS 2025 Workshop Paper | Lyle Ungar

Comparing Collective Behavior of LLM and Human Groups
12/6: NeurIPS 2025 Workshop Paper | Chris Callison-Burch

Learning Human-Perceived Fakeness in AI-Generated Videos via Multimodal LLMs
12/6: NeurIPS 2025 Workshop Paper | Chris Callison-Burch

Probabilistic Soundness Guarantees in LLM Reasoning Chains
12/6: NeurIPS 2025 Workshop Paper | Chris Callison-Burch

In Good GRACEs: Principled Teacher Selection for Knowledge Distillation
12/6: NeurIPS 2025 Workshop Paper | Surbhi Goel

Optimal Lower Bounds and New Upper Bounds for Sequential Prediction With Abstention
12/6: NeurIPS 2025 Workshop Paper | Surbhi Goel

Reliable Human-AI Collaboration via Agreement and Competition
12/6: NeurIPS 2025 Workshop Talk | Surbhi Goel

Testing Noise Assumptions of Learning Algorithms
12/6: NeurIPS 2025 Workshop Paper | Surbhi Goel

Alignment via Competition: Emergent Alignment From Differently Misaligned Agents
12/6: NeurIPS 2025 Workshop Paper | Surbhi Goel and Aaron Roth

Collaborative Prediction: Tractable Information Aggregation via Agreement
12/6: NeurIPS 2025 Workshop Paper | Surbhi Goel and Aaron Roth

Benchmarking Mitigations Against Covert Misuse
12/6: NeurIPS 2025 Workshop Paper | Hamed Hassani

BrowserArena: Evaluating LLM Agents on Real-World Web Navigation Tasks
12/6: NeurIPS 2025 Workshop Paper | Hamed Hassani

High-Resolution Weak Lensing Mass Mapping From DES-Y3 Data Using Diffusion-Based Prior
12/6: NeurIPS 2025 Workshop Paper | Bhuvnesh Jain

Vision Transformers for Cosmological Fields: Application to Weak Lensing k Maps
12/6: NeurIPS 2025 Workshop Paper | Bhuvnesh Jain

Algorithmic Aspects of Strategic Trading
12/6: NeurIPS 2025 Workshop Paper | Michael Kearns

Replicable Reinforcement Learning With Linear Function Approximation
12/6: NeurIPS 2025 Workshop Paper | Michael Kearns and Aaron Roth

PrimateFace: A Resource for Generalizable Cross-Species Facial Analysis
12/6: NeurIPS 2025 Workshop Paper | Konrad Kording

Long-Term Wireless Link Scheduling With State-Augmented Graph Neural Networks
12/6: NeurIPS 2025 Workshop Poster | Alejandro Ribeiro

Statistical Early Stopping for Reasoning Models
12/6: NeurIPS 2025 Workshop Paper | Weijie Su and Edgar Dobriban

Efficient Reinforcement Learning for Optimizing Multi-Turn Student Outcomes With LLM Tutors
12/6: NeurIPS 2025 Workshop Paper | Lyle Ungar

Conformal Information Pursuit for Interactively Guiding Large Language Models
12/5: NeurIPS 2025 Poster | Edgar Dobriban, Hamed Hassani, and René Vidal

On the Mechanisms of Weak-to-Strong Generalization: A Theoretical Perspective
12/5: NeurIPS 2025 Poster | Hamed Hassani

A Scalable, Causal, and Energy Efficient Framework for Neural Decoding With Spiking Neural Networks
12/5: NeurIPS 2025 Poster | Konrad Kording

SECA: Semantically Equivalent and Coherent Attacks for Eliciting LLM Hallucinations
12/5: NeurIPS 2025 Poster | René Vidal

Foundations of Top-k Decoding for Language Models
12/4: NeurIPS 2025 Poster and Workshop Paper | Edgar Dobriban

Synthetic-Powered Predictive Inference
12/4: NeurIPS 2025 Poster | Edgar Dobriban

Optimal Neural Compressors for the Rate-Distortion-Perception Tradeoff
12/4: NeurIPS 2025 Spotlight Poster and Workshop Paper | Hamed Hassani and Shirin Saeedi Bidokhti

A Dataset for Distilling Knowledge Priors From Literature for Therapeutic Design
12/4: NeurIPS 2025 Poster | Zack Ives and Yoseph Barash

A Statistical Framework of Watermarks for Large Language Models: Pivot, Detection Efficiency and Optimal Rules
12/4: NeurIPS 2025 Poster | Weijie Su

Does Object Binding Naturally Emerge in Large Pretrained Vision Transformers?
12/4: NeurIPS 2025 Spotlight Poster | Lyle Ungar and Konrad Kording

Neural Collapse Under Gradient Flow on Shallow ReLU Networks for Orthogonally Separable Data
12/4: NeurIPS 2025 Poster | René Vidal

Efficient PAC Learning for Realizable-Statistic Models via Convex Surrogates
12/3: NeurIPS 2025 Poster | Shivani Agarwal

Conformal Inference Under High-Dimensional Covariate Shifts via Likelihood-Ratio Regularization
12/3: NeurIPS 2025 Poster | Edgar Dobriban and Hamed Hassani

Probabilistic Stability Guarantees for Feature Attributions
12/3: NeurIPS 2025 Poster | Surbhi Goel

Learning Theory Alliance
12/3: NeurIPS 2025 Social | Surbhi Goel, Shivani Agarwal, and Hamed Hassani

Conformal Prediction Beyond the Seen: A Missing Mass Perspective for Uncertainty Quantification in Generative Models
12/3: NeurIPS 2025 Poster | Hamed Hassani

Alignment of Large Language Models With Constrained Learning
12/3: NeurIPS 2025 Poster and Workshop Paper | Alejandro Ribeiro

Composition and Alignment of Diffusion Models Using Constrained Learning
12/3: NeurIPS 2025 Poster and Workshop Paper | Alejandro Ribeiro

Mitigating the Privacy–Utility Trade-Off in Decentralized Federated Learning via f-Differential Privacy
12/3: NeurIPS 2025 Spotlight Poster | Weijie Su

On the Empirical Power of Goodness-of-Fit Tests in Watermark Detection
12/3: NeurIPS 2025 Spotlight Poster | Weijie Su

DiverseDialogue: A Methodology for Designing Chatbots With Human-Like Diversity
12/1: NeurIPS 2025 Workshop Paper | Lyle Ungar

Convergence Rates for Gradient Descent on the Edge of Stability for Overparametrised Least Squares
12/3: NeurIPS 2025 Poster | René Vidal

MotionBind: Multi-Modal Human Motion Alignment for Retrieval, Recognition, and Generation
12/3: NeurIPS 2025 Poster | René Vidal

Website

Get Involved

Past Events

107 Towne Building
220 South 33rd Street
Philadelphia, Pennsylvania 19104
warrenct@seas.upenn.edu