Thomas Schmied
Papers
1
Total Citations
2
H-Index
1
About
Thomas Schmied is a leading researcher at the intersection of reinforcement learning (RL) and sequence modeling, with a focus on developing efficient, scalable architectures for robotics. His most-cited work, "A Large Recurrent Action Model: xLSTM enables Fast Inference for Robotics Tasks" (2024), introduces a novel approach that leverages extended Long Short-Term Memory (xLSTM) networks to replace traditional Transformer-based models in offline RL. This contribution addresses a critical bottleneck: while Transformers have enabled powerful large action models, their slow inference limits real-time robotic applications. Schmied’s xLSTM-based architecture achieves comparable performance with significantly faster inference, making it ideal for deployment in latency-sensitive environments. With 2 citations in its first year, this work is gaining traction for its practical impact. Schmied’s research advances the frontier of efficient, offline-trained agents, bridging the gap between large-scale sequence modeling and real-world robotics. His work is particularly notable for challenging the dominance of Transformers in RL, offering a compelling alternative that prioritizes speed without sacrificing capability.
Research Focus
Key Achievements
Top Papers
- 1