MidSteer: Optimal Affine Framework for Steering Generative Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918532114022400 |
|---|---|
| author | Gaintseva, Tatiana Stepanov, Andrew Liu, Ziquan Benning, Martin Slabaugh, Gregory Deng, Jiankang Elezi, Ismail |
| author_facet | Gaintseva, Tatiana Stepanov, Andrew Liu, Ziquan Benning, Martin Slabaugh, Gregory Deng, Jiankang Elezi, Ismail |
| contents | Steering intermediate representations has emerged as a powerful strategy for controlling generative models, particularly in post-deployment alignment and safety settings. However, despite its empirical success, it currently lacks a comprehensive theoretical framework. In this paper, we bridge this gap by formalizing the theory of concept steering. First, we establish a link between steering and affine concept erasure, proving that the standard approach for removing unwanted behaviors is a special case of LEACE (a closed-form method for affine erasure). Next, we formulate a principled theoretical framework for concept switching, LEACE-Switch, and characterize the assumptions under which it provides an optimal affine solution. Building on this analysis, we then introduce MidSteer (Minimal Disturbance concept Steering), a more general affine framework for concept manipulation that relaxes these assumptions and enables directed, minimal-disturbance transformations. We demonstrate that MidSteer performs favorably across a range of tasks, modalities, and architectures, including vision diffusion models and large language models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_05220 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MidSteer: Optimal Affine Framework for Steering Generative Models Gaintseva, Tatiana Stepanov, Andrew Liu, Ziquan Benning, Martin Slabaugh, Gregory Deng, Jiankang Elezi, Ismail Machine Learning Artificial Intelligence Steering intermediate representations has emerged as a powerful strategy for controlling generative models, particularly in post-deployment alignment and safety settings. However, despite its empirical success, it currently lacks a comprehensive theoretical framework. In this paper, we bridge this gap by formalizing the theory of concept steering. First, we establish a link between steering and affine concept erasure, proving that the standard approach for removing unwanted behaviors is a special case of LEACE (a closed-form method for affine erasure). Next, we formulate a principled theoretical framework for concept switching, LEACE-Switch, and characterize the assumptions under which it provides an optimal affine solution. Building on this analysis, we then introduce MidSteer (Minimal Disturbance concept Steering), a more general affine framework for concept manipulation that relaxes these assumptions and enables directed, minimal-disturbance transformations. We demonstrate that MidSteer performs favorably across a range of tasks, modalities, and architectures, including vision diffusion models and large language models. |
| title | MidSteer: Optimal Affine Framework for Steering Generative Models |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2605.05220 |