3 papers
cs.CV2025
When Preferences Diverge: Aligning Diffusion Models with Minority-Aware Adaptive DPO
Lingfan Zhang, Chen Liu, Chengming Xu +5
In recent years, the field of image generation has witnessed significant advancements, particularly in fine-tuning methods that align models with universal human preferences. This…
cs.LG2024
Mitigating the Alignment Tax of RLHF
Yong Lin, Hangyu Lin, Wei Xiong +14
LLMs acquire a wide range of abilities during pre-training, but aligning LLMs under Reinforcement Learning with Human Feedback (RLHF) can lead to forgetting pretrained abilities, w…
cs.LG2024
A Generalization Theory of Cross-Modality Distillation with Contrastive Learning
Hangyu Lin, Chen Liu, Chengming Xu +3
Cross-modality distillation arises as an important topic for data modalities containing limited knowledge such as depth maps and high-quality sketches. Such techniques are of great…