Representation Engineering: A Top-Down Approach to AI Transparency
by Center for AI Safety / UCSD · free · Last verified 2026-03-17
Representation Engineering (RepE) is a top-down AI transparency technique that identifies and manipulates high-level concepts within a model's activations. By finding linear directions corresponding to traits like honesty or power-seeking, it enables real-time monitoring and steering of model behavior, offering a scalable alternative to circuit-level analysis.
https://arxiv.org/abs/2310.01405 ↗B
B—Above Average
Adoption: BQuality: AFreshness: B+Citations: B+Engagement: F
Specifications
- License
- Open Access
- Pricing
- free
- Capabilities
- Concept Vector Identification, Model Behavior Steering, Real-time Behavior Monitoring, Scalable Interpretability Analysis, Honesty and Safety Control, Bias and Emotion Detection, Linear Representation Probing, Activation Space Manipulation
- Integrations
- Use Cases
- [object Object], [object Object], [object Object], [object Object]
- API Available
- No
- Tags
- interpretability, representation-engineering, transparency, control-vectors, llm, ai-safety, model-alignment, activation-engineering, concept-vectors, explainable-ai
- Added
- 2026-03-17
- Completeness
- 0.9%
Index Score
61.6Adoption
65
Quality
88
Freshness
78
Citations
72
Engagement
0