Princeton University
Harvard University
Robert LasenbyMember of Technical Staff · AnthropicResearcher↗
Researcher
Samuel Marks
AnthropicScalable Oversight Lead
Samuel Marks develops ways to uncover hidden behavior in AI models. He co-led an experiment that tested whether auditors could discover a goal deliberately hidden in a model.[1] He also helped build automated tools used to audit Claude 4.[2] His Sparse Feature Circuits work traced model behavior to patterns inside a network.[3]
Career & education
Current roleAnthropic2024-presentScalable Oversight Lead
Anthropic2024-presentPreviously
Northeastern University2023-2024Postdoctoral Researcher
Northeastern University2023-2024Work and education history is primarily focused on AI-relevant roles and may not be comprehensive.
Citation Trend
Total citations to scholarly work, including research papers, technical reports, books and theses, as recorded by Google Scholar. Each point is a dated snapshot.
citations7,094As of Nov 14, '26
References
- Auditing language models for hidden objectivesanthropic.com
Marks is a first-listed equal contributor on the linked paper, which tests whether auditors can discover an intentionally hidden objective.
- Building and evaluating alignment auditing agentsalignment.anthropic.com
Credits Marks and describes using automated agents to assist with Claude 4 audits.
- Sparse Feature Circuitsarxiv.org
Marks is first-listed on the paper introducing a method to trace behavior to interpretable features.
Based on public sources. If accuracy is important, verify details using the links or references.
For corrections, updates, or requests to remove information, contact us.
