Mehryar Mohri
4 papers · Latest:
Generalized Distributional Alignment Games for Unbiased Answer-Level Fine-Tuning
This paper resolves systematic estimation bias in Distributional Alignment Games for Answer-Level Fine-Tuning, leading to more stable and efficient training.
Linear-Core Surrogates: Smooth Loss Functions with Linear Rates for Classification and Structured Prediction
Linear-Core (LC) Surrogates are new smooth loss functions offering fast optimization and linear consistency rates for classification and structured prediction.
Mind the Gap: Structure-Aware Consistency in Preference Learning
This paper introduces SA-DPO, a new preference learning method for LLMs that ensures theoretical consistency and adapts margins based on semantic distance.
Optimized Deferral for Imbalanced Settings
MILD is a new framework addressing expert imbalance in two-stage learning to defer by using cost-sensitive learning and novel margin-based loss functions.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.