Hi there!😊 I’m Kairan Dou, welcome to my personal website! Feel free to just call me Kevin.

I am a junior undergraduate student majoring in Computer Science at Nankai University, and I have just completed my exchange program at the University of California, Berkeley.

This summer, I am working as a research intern at the both MIT Media Lab and Harvard Ophthalmology AI Lab under the joint supervision of Prof.Paul Liang and Prof.Mengyu Wang, where I focus on robotic manipulation. Since February 2025, I have been a Research Assistant at The University of Texas at Austin, working under the guidance of Prof. Philipp Krähenbühl, focusing on multimodal learning. Previously, I have conducted extensive research at the Visual Computing and Intelligent Perception(VCIP) Lab, advised by Prof. Xiang Li.

My current research interests lie in:

Reinforcement learning for stability and alignment in VLA models
Enhancing few-shot and zero-shot generalization in embodied agents

I aim to develop embodied agents with the capacity for generalizable reasoning and long-horizon decision-making. My long-term vision is to advance the foundations of real-world intelligence through unified perception, control, and learning.

I am planning to apply for PhD programs in Fall 2026, with a research focus on reinforcement learning, multimodal reasoning, and embodied AI. If you are recruiting or open to collaboration, I would be glad to connect:)

You can also reach me on WeChat at: Darkeyes-

🔥 News

2025.05: 🎉🎉 Our paper was accepted at FMEA Workshop @ CVPR 2025.
2025.02: 🎉🎉 I delivered an oral presentation at AAAI 2025 in Philadelphia.
2024.12: 🎉🎉 Our paper was accepted at AAAI 2025.

📝 Publications

FMEA Workshop

Interactive Post-Training for Vision-Language-Action Models

Shuhan Tan, Kairan Dou, Yue Zhao, Philipp Krähenbüh

Introduces RIPT-VLA, a scalable third-stage reinforcement learning method for VLA models, enhancing performance through interactive training with sparse binary rewards.
Achieves SOTA performance across diverse benchmarks, including LIBERO-90 (94.3%), LIBERO-LONG 5-shot (71.4%), MetaWorld45 5-shot (76.0%), and OpenVLA-OFT (97.5%).
Employs dynamic rollout sampling and leave-one-out advantage estimation to significantly enhance generalization, stability, and effectiveness across challenging tasks and scenarios.

AAAI 2025

From Words to Worth: Newborn Article Impact Prediction with LLM

Penghai Zhao, Xinghua Xing, Kairan Dou, Jinyu Tian, Ying Tai, Jian Yang, Ming-Ming Cheng, Xiang Li

Proposed the “Newborn Article Impact Prediction” (Newborn AIP) task and introduced the TNCSIsp metric, achieving an NDCG@20 score of 0.901.
Constructed TKPD and NAID datasets, including over 12,000 samples for training and validation.
Used LoRA to fine-tune and test 5+ large language models on server to evaluate prediction performance.