Scholar
Samuel Holt
Google Scholar ID: Ey5aInIAAAAJ
University of Cambridge
Citations & Impact
Citations
401
H-index
12
i10-index
13
Publications
20
Co-authors
0
Publications
Cited
0
Resume
Academic Achievements
- Published twelve papers as first or joint first author in top-tier ML conferences (e.g., NeurIPS [spotlight], ICML [long oral], ICLR [spotlight], RSS, AISTATS). Specific works include improving LLM agent planning with in-context learning and introducing EvoControl: Multi-Frequency Bi-Level Control for High-Frequency Continuous Control.
Research Experience
- Recent RS intern at Google DeepMind and a fourth-year Ph.D. student in Machine Learning at the University of Cambridge, part of the Machine Learning and Artificial Intelligence group. Involved in multiple LLM-related research projects.
Education
- PhD in Machine Learning, 2021 - 2025, University of Cambridge, advised by Mihaela van der Schaar FRS; MEng in Engineering Science, 2013 - 2017, University of Oxford.
Background
- Interests include Large Language Models (LLMs), LLM Agents, Transformer Architectures, Reinforcement Learning (model-free and model-based), Control, Symbolic Regression (discovery), and applications to scientific discovery and healthcare. Focuses on driving foundational research to advance the state-of-the-art for LLM output generation, particularly in LLM agents using external memory and tool use.
Miscellany
- Passionate about inventing and flexibly adapting to drive new prototypes forward, especially in areas like multi-modal LLM agents and using RL to automatically improve LLM agents.