CircuitSteer: Geometrically Aligned Multi-Layer Steering via Sparse Autoencoder Circuits
Controlling the behavior of large language models (LLMs) remains a critical challenge for AI alignment. Existing steering methods, such as Contrastive Activation Addition (CAA), typically rely on fixed single-layer interventions derived from aggregate activation differences. These methods impose a single intervention across semantically diverse inputs and often fail to sustain consistent behavioral changes across layers, limiting the effectiveness of the steering. In this work, we introduce Circ
Record details
Published: 6 August 2026
Source: arXiv alignment query
Category: Research
Topics: Safety & alignment
Retrieved: 7 August 2026
Related evidence
These records share source-supplied organisations, an exact publisher byline, automatic topics or regions. The reason is shown on every link; related does not mean supporting, agreeing with or verifying this record.
DistMedVL: Distributional Vision-Language Alignment for Uncertainty-Aware Medical Image Segmentation
arXiv · 6 August 2026
Breaking Customized LLMs for Coding: Automated Red Teaming for Instruction Backdoor Attacks
arXiv red teaming query · 6 August 2026
Studying People to Study AI: Expert Perspectives on the Epistemic Fit and Barriers of Human Research in AI Safety & Ethics
arXiv cs.HC · 6 August 2026
Studying People to Study AI: Expert Perspectives on the Epistemic Fit and Barriers of Human Research in AI Safety & Ethics
arXiv · 6 August 2026
Functional Misalignment in Human-AI Interactions on Digital Platforms
arXiv cs.CY · 6 August 2026
How Should AI Safety Benchmarks Benchmark Safety?
arXiv cs.CY · 6 August 2026
How to cite this record
ethics.ai (6 August 2026), “CircuitSteer: Geometrically Aligned Multi-Layer Steering via Sparse Autoencoder Circuits,” evidence record 17386, https://ethics.ai/record/17386 (originally published by arXiv alignment query).
Use and limitations
This page is a stable index and citation surface for a source record. ethics.ai did not author the underlying report and has not independently verified every claim. Automatic topics may be imperfect. For consequential use, quote and cite the original publisher.