We Need A Science of Scheming
Advanced AI systems face strong incentives to scheme. Apollo Research is building a science of scheming to predict and prevent this risk.
19/01/2026
Read more
Science
We conduct fundamental research into the science of scheming and its potential mitigations, as well as develop and run pre-deployment evaluations of frontier AI systems.
our key papers
All posts
Filter
- All
- Science of Scheming
- Evaluations
- Interpretability
- Notes
Science of Scheming
Stress Testing Deliberative Alignment for Anti-Scheming Training
17/09/2025
Read more
Evaluations
Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety
15/07/2025
Read more
Evaluations
Research Note: Our scheming precursor evals had limited predictive power for our in-context scheming evals
03/07/2025
Read more
Interpretability
Interpretability in Parameter Space: Minimizing Mechanistic Description Length with Attribution-based Parameter Decomposition
11/02/2025
Read more
Interpretability
Identifying functionally important features with end-to-end sparse dictionary learning
30/05/2024
Read more
Interpretability
The Local Interaction Basis: Identifying Computationally-Relevant and Sparsely Interacting Features in Neural Networks
30/05/2024
Read more
Evaluations
Large Language Models can Strategically Deceive their Users when Put Under Pressure
09/11/2023
Read more
