Gaotang Li

Beyond Log Likelihood: Probability-Based Objectives for Supervised Fine-Tuning across the Model Capability Continuum

RM-R1: Reward Modeling as Reasoning

Graph homophily booster: Reimagining the role of discrete features in heterophilic graph learning

Stabilizing Reinforcement Learning for Honesty Alignment in Language Models on Deductive Reasoning

Moralise: A Structured Benchmark for Moral Alignment in Visual Language Models

Taming Knowledge Conflicts in Language Models