Papers
3
Total Citations
15
H-Index
2
About
Kai Lv is a rising researcher in artificial intelligence, whose work is shaping the future of safe and generalizable reinforcement learning (RL). His primary research areas include visual RL generalization, constrained RL, and multi-objective decision-making for robotics. Lv’s most impactful contribution is a pioneering theoretical analysis of generalization in visual RL, establishing an upper bound on the generalization objective that links policy divergence and Bellman error. This work, published in 2024 and already garnering 9 citations, provides a foundational framework for building RL agents that can reliably adapt to unseen environments. He has also advanced the critical field of safe RL by developing an adaptive ensemble C estimation method for off-policy constrained RL, addressing the challenge of enforcing safety constraints in real-world agents. Most recently, Lv has extended these principles to complex robotic systems, tackling multi-constraint scenarios. His research is notable for bridging rigorous theory with practical deployment challenges, making him a key voice in the effort to create robust, trustworthy AI agents for real-world interaction.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3Multi-constraint reinforcement learning in complex robot environments2 citations · 2025