316 citations · 661 across the 65 of their papers we have counts for
70 papers
An Efficient and Modular Framework for Targeted Harm Mitigation in LLMS
Roberto Campbell, Momin Abbas, Momin Abbass +6
Large Language Models (LLMs) are powerful zero-shot learners but remain prone to misalignment with human preferences, often producing biased, toxic, or otherwise harmful outputs. E…
Stop Guessing When to Stop Testing: Efficient Model Evaluation with Just Enough Data
Ofir Arviv, Kristjan Greenewald, Yotam Perlitz +3
The inherent rigidity of fixed-size benchmarks makes them an inefficient tool for model evaluation. Diverse evaluation objectives, including model ranking, model selection and test…
Distributional Process Reward Models: Calibrated Prediction of Future Rewards via Conditional Optimal Transport
Rachel Ma, Dylan Hadfield-Menell, Kristjan Greenewald
Inference-time scaling methods rely on Process Reward Models (PRMs), which are often poorly calibrated and overestimate success probabilities. We propose, to our knowledge, the fir…
Efficient Multi-Adapter LLM Serving via Cross-Model KV-Cache Reuse with Activated LoRA
Allison Li, Kristjan Greenewald, Thomas Parnell +1
Modern large language model (LLM) systems increasingly rely on multi-turn pipelines that are composed of multiple task-specific adapters, yet existing serving frameworks remain ine…
Differentially Private Wasserstein Barycenters
Anming Gu, Sasidhar Kunapuli, Mark Bun +2
The Wasserstein barycenter is defined as the mean of a set of probability measures under the optimal transport metric, and has numerous applications spanning machine learning, stat…
Entropic Causal Inference: Graph Identifiability
Spencer Compton, Kristjan Greenewald, Dmitriy Katz +1
Entropic causal inference is a recent framework for learning the causal graph between two variables from observational data by finding the information-theoretically simplest struct…
Private Continuous-Time Synthetic Trajectory Generation via Mean-Field Langevin Dynamics
Anming Gu, Edward Chien, Kristjan Greenewald
We provide an algorithm to privately generate continuous-time data (e.g. marginals from stochastic differential equations), which has applications in highly sensitive domains invol…
Know What You Don't Know: Uncertainty Calibration of Process Reward Models
Young-Jin Park, Kristjan Greenewald, Kaveh Alim +2
Process reward models (PRMs) play a central role in guiding inference-time scaling algorithms for large language models (LLMs). However, we observe that even state-of-the-art PRMs…
Neural Estimation for Scaling Entropic Multimarginal Optimal Transport
Dor Tsur, Ziv Goldfeld, Kristjan Greenewald +1
Multimarginal optimal transport (MOT) is a powerful framework for modeling interactions between multiple distributions, yet its applicability is bottlenecked by a high computationa…
Revisiting Group Relative Policy Optimization: Insights into On-Policy and Off-Policy Training
Youssef Mroueh, Nicolas Dupuis, Brian Belgodere +6
We revisit Group Relative Policy Optimization (GRPO) in both on-policy and off-policy optimization regimes. Our motivation comes from recent work on off-policy Proximal Policy Opti…
Fairness Is More Than Algorithms: Racial Disparities in Time-to-Recidivism
Jessy Xinyi Han, Kristjan Greenewald, Devavrat Shah
Racial disparities in recidivism remain a persistent challenge, and the growing adoption of risk assessment algorithms has intensified the scrutiny of their sources. Past works hav…
Activated LoRA: Fine-tuned LLMs for Intrinsics
Kristjan Greenewald, Luis Lastras, Thomas Parnell +6
Low-Rank Adaptation (LoRA) has emerged as a highly efficient framework for finetuning the weights of large foundation models, and has become the go-to method for data-driven custom…
A Library of LLM Intrinsics for Retrieval-Augmented Generation
Marina Danilevsky, Kristjan Greenewald, Chulaka Gunasekara +13
In the developer community for large language models (LLMs), there is not yet a clean pattern analogous to a software library, to support very large scale collaboration. Even for t…
Verify when Uncertain: Beyond Self-Consistency in Black Box Hallucination Detection
Yihao Xue, Kristjan Greenewald, Youssef Mroueh +1
Large Language Models (LLMs) often hallucinate, limiting their reliability in sensitive applications. In black-box settings, several self-consistency-based techniques have been pro…
Privacy without Noisy Gradients: Slicing Mechanism for Generative Model Training
Kristjan Greenewald, Yuancheng Yu, Hao Wang +1
Training generative models with differential privacy (DP) typically involves injecting noise into gradient updates or adapting the discriminator's training procedure. As a result,…
Domain Adaptable Prescriptive AI Agent for Enterprise
Piero Orderique, Wei Sun, Kristjan Greenewald
Despite advancements in causal inference and prescriptive AI, its adoption in enterprise settings remains hindered primarily due to its technical complexity. Many users lack the ne…
Gradient Flows and Riemannian Structure in the Gromov-Wasserstein Geometry
Zhengxin Zhang, Ziv Goldfeld, Kristjan Greenewald +2
The Wasserstein space of probability measures is known for its intricate Riemannian structure, which underpins the Wasserstein geometry and enables gradient flow algorithms. Howeve…
Compress then Serve: Serving Thousands of LoRA Adapters with Little Overhead
Rickard Brüel-Gabrielsson, Jiacheng Zhu, Onkar Bhardwaj +4
Fine-tuning large language models (LLMs) with low-rank adaptations (LoRAs) has become common practice, often yielding numerous copies of the same LLM differing only in their LoRA u…
Multivariate Stochastic Dominance via Optimal Transport and Applications to Models Benchmarking
Gabriel Rioux, Apoorva Nitsure, Mattia Rigotti +2
Stochastic dominance is an important concept in probability theory, econometrics and social choice theory for robustly modeling agents' preferences between random outcomes. While m…
Distributional Preference Alignment of LLMs via Optimal Transport
Igor Melnyk, Youssef Mroueh, Brian Belgodere +6
Current LLM alignment techniques use pairwise human preferences at a sample level, and as such, they do not imply an alignment on the distributional level. We propose in this paper…
Slicing Mutual Information Generalization Bounds for Neural Networks
Kimia Nadjahi, Kristjan Greenewald, Rickard Brüel Gabrielsson +1
The ability of machine learning (ML) algorithms to generalize well to unseen data has been studied through the lens of information theory, by bounding the generalization error with…
Partially Observed Trajectory Inference using Optimal Transport and a Dynamics Prior
Anming Gu, Edward Chien, Kristjan Greenewald
Trajectory inference seeks to recover the temporal dynamics of a population from snapshots of its (uncoupled) temporal marginals, i.e. where observed particles are not tracked over…
Score Distillation via Reparametrized DDIM
Artem Lukoianov, Haitz Sáez de Ocáriz Borde, Kristjan Greenewald +4
While 2D diffusion models generate realistic, high-detail images, 3D shape generation methods like Score Distillation Sampling (SDS) built on these 2D diffusion models produce cart…
Synthetic Census Data Generation via Multidimensional Multiset Sum
Cynthia Dwork, Kristjan Greenewald, Manish Raghavan
The US Decennial Census provides valuable data for both research and policy purposes. Census data are subject to a variety of disclosure avoidance techniques prior to release in or…