Yixuan Li

Examining how data shapes the alignment of AI models with human preferences.

Introducing a new framework for realistic pedestrian animations in AV training.

Introducing methods to guide language models in displaying specific personality traits.

This article discusses the importance and challenges of out-of-distribution detection in machine learning.

Examining how in-distribution labels impact out-of-distribution detection in machine learning.

Current models struggle with spatial reasoning, relying more on text than images.

Exploring the interplay of quiver Hecke algebras and Floer homology in modern mathematics.

A new framework improves alignment of language models with user preferences.

R2G improves robots' ability to follow verbal instructions in 3D spaces.

INK provides a reliable method for identifying out-of-distribution samples in machine learning.

A new method integrates human input to enhance OOD learning for machine learning models.

Using weaker language models can improve AI alignment efficiently.

A new framework improves detection of false outputs in language models using unlabeled data.

A graph-based approach to enhance machine learning in dynamic environments.

CodeLutra teaches models to learn from their successes and failures.

Learn how CONDA helps AI adapt and remain interpretable in changing conditions.

Evaluating AI images to ensure effective communication in advertising.