Xinran Liang

I am a PhD student in Computer Science at Princeton University, advised by Chi Jin.

Previously, I received my bachelor degree in Applied Mathematics and Data Science from UC Berkeley. I did research as part of Berkeley Artificial Intelligence Research, where I was advised by Kimin Lee, Aditi Raghunathan, and Pieter Abbeel.

Email / Resume / Google Scholar / Github

Research

My research focuses on multimodal generative models, such as vision language models and visual synthesis models, and applications of deep reinforcement learning algorithms, with the goal of improving reasoning, decision-making, and generalization capabilities of multimodal models.

	Odysseus: Scaling VLMs to 100+ Turn Decision-Making in Games via Reinforcement Learning Chengshuai Shi, Wenzhe Li, Xinran Liang, Yizhou Lu, Wenjia Yang, Ruirong Feng, Seth Karten, Ziran Yang, Zihan Ding, Gabriel Sarch, Danqi Chen, Karthik Narasimhan, Chi Jin arxiv preprint*, 2026 paper / website We introduce Odysseus, an open training framework for reinforcement learning of vision-language models on 100+ turn decision-making in visually grounded games.
	Personalized Generative Models for Contextual Debiasing Xinran Liang, Esin Tureci, Prachi Sinha, Ye Zhu, Vikram Ramaswamy, Olga Russakovsky CVPR Workshop on Synthetic Data for Computer Vision, 2026 paper / code We introduce a method to decouple contextual patterns in vision datasets: it personalizes text-to-image diffusion models to synthesize training augmentations with uncommon visual contexts while preserving alignment with the original dataset.
	ALP: Action-Aware Embodied Learning for Perception Xinran Liang, Anthony Han, Wilson Yan, Aditi Raghunathan, Pieter Abbeel arxiv preprint, 2023 paper / website / code An embodied learning framework based on active exploration for visual representations and perception tasks. We propose to learn representations from action signals implicitly through reinforcement learning and explicitly via inverse dynamics prediction.
	Reward Uncertainty for Exploration in Preference-based Reinforcement Learning Xinran Liang, Katherine Shu, Kimin Lee, Pieter Abbeel International Conference on Learning Representations (ICLR), 2022 paper / code We propose a simple and efficient human-guided exploration method by measuring uncertainty in human instructions as intrinsic rewards.

Honors and Awards

Department Citation, Department of Data Science, Uiversity of California Berkeley, 2022
Outstanding Graduate Student Instructor Award, Uiversity of California Berkeley, 2021-2022
Percy Lionel Davis Award, Department of Mathematics, Uiversity of California Berkeley, 2022

Teaching

CS 324: Introduction to Machine Learning
Graduate Student Instructor: Fall 2023

CS 226: Algorithms and Data Structures
Graduate Student Instructor: Spring 2024

Data 140: Probability for Data Science
Head Undergraduate Student Instructor: Spring 2022, Fall 2021, Spring 2021
Undergraduate Student Instructor: Fall 2020
Group Tutor: Spring 2020

Data 100: Principles and Techniques of Data Science
Undergraduate Student Instructor: Summer 2020

Redesigned from Jon Barron's source code.