Daniel Geng
Papers
2
Total Citations
29
H-Index
2
About
Daniel Geng is a rising researcher whose work explores the intersection of reinforcement learning, intrinsic motivation, and emergent behavior in artificial agents. His most influential contribution is the development of **SMiRL (Surprise Minimizing Reinforcement Learning)**, a paradigm that reframes agency as a struggle against environmental disorder. In his highly cited 2019 papers—garnering 17 and 12 citations respectively—Geng proposes that agents can learn useful, robust behaviors simply by minimizing surprise, or the divergence between predicted and actual sensory states. This unsupervised principle allows agents to carve out stable niches in dynamic, unstable environments without explicit reward signals. By drawing inspiration from biological homeostasis, Geng’s work offers a compelling alternative to traditional reward-based learning, showing how order-seeking behavior can emerge naturally. His research has significant implications for robotics, game AI, and autonomous systems operating in unpredictable real-world conditions. Geng’s SMiRL framework stands as a notable achievement, demonstrating how simple, biologically-inspired rules can lead to complex, adaptive intelligence—a concept that continues to influence ongoing work in intrinsic motivation and open-ended learning.
Research Focus
Key Achievements
Top Papers
- 1SMiRL: Surprise Minimizing RL in Dynamic Environments17 citations · 2019
- 2SMiRL: Surprise Minimizing Reinforcement Learning in Unstable Environments12 citations · 2019