Shih-Wei Liao
Papers
3
Total Citations
49
H-Index
3
About
Shih-Wei Liao is a leading researcher in Multi-Agent Reinforcement Learning (MARL), with a sharp focus on the algorithmic foundations that make cooperative AI systems reliable and efficient. His work critically examines the widely used QMIX algorithm, particularly its monotonicity constraint and the often-overlooked implementation tricks that can dramatically affect performance. In his highly cited 2021 paper, "Rethinking the Implementation Tricks and Monotonicity Constraint in Cooperative Multi-Agent Reinforcement Learning" (35 citations), Liao systematically dissects the components of state-of-the-art MARL algorithms, revealing how subtle coding choices can lead to significant performance gains or failures. His follow-up studies, "RIIT" (10 citations) and "Revisiting the Monotonicity Constraint" (4 citations), further challenge conventional wisdom by questioning the necessity and impact of the monotonicity constraint itself. By demystifying these technical nuances, Liao’s work provides essential guidance for researchers and engineers building scalable multi-agent systems, from robot swarms to autonomous vehicle coordination. His contributions are a must-read for anyone seeking to move beyond black-box implementations and truly understand what makes cooperative MARL work.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3