SAE-SSV: Supervised Steering in Sparse Representation Spaces for Reliable Control of Language Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914180915789824 |
|---|---|
| author | He, Zirui Jin, Mingyu Shen, Bo Payani, Ali Zhang, Yongfeng Du, Mengnan |
| author_facet | He, Zirui Jin, Mingyu Shen, Bo Payani, Ali Zhang, Yongfeng Du, Mengnan |
| contents | Large language models (LLMs) have demonstrated impressive capabilities in natural language understanding and generation, but controlling their behavior reliably remains challenging, especially in open-ended generation settings. This paper introduces a novel supervised steering approach that operates in sparse, interpretable representation spaces. We employ sparse autoencoders (SAEs) to obtain sparse latent representations that aim to disentangle semantic attributes from model activations. Then we train linear classifiers to identify a small subspace of task-relevant dimensions in latent representations. Finally, we learn supervised steering vectors constrained to this subspace, optimized to align with target behaviors. Experiments across sentiment, truthfulness, and political polarity steering tasks with multiple LLMs demonstrate that our supervised steering vectors achieve higher success rates with minimal degradation in generation quality compared to existing methods. Further analysis reveals that a notably small subspace is sufficient for effective steering, enabling more targeted and interpretable interventions. Our implementation is publicly available at https://github.com/Ineedanamehere/SAE-SSV. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_16188 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SAE-SSV: Supervised Steering in Sparse Representation Spaces for Reliable Control of Language Models He, Zirui Jin, Mingyu Shen, Bo Payani, Ali Zhang, Yongfeng Du, Mengnan Computation and Language Large language models (LLMs) have demonstrated impressive capabilities in natural language understanding and generation, but controlling their behavior reliably remains challenging, especially in open-ended generation settings. This paper introduces a novel supervised steering approach that operates in sparse, interpretable representation spaces. We employ sparse autoencoders (SAEs) to obtain sparse latent representations that aim to disentangle semantic attributes from model activations. Then we train linear classifiers to identify a small subspace of task-relevant dimensions in latent representations. Finally, we learn supervised steering vectors constrained to this subspace, optimized to align with target behaviors. Experiments across sentiment, truthfulness, and political polarity steering tasks with multiple LLMs demonstrate that our supervised steering vectors achieve higher success rates with minimal degradation in generation quality compared to existing methods. Further analysis reveals that a notably small subspace is sufficient for effective steering, enabling more targeted and interpretable interventions. Our implementation is publicly available at https://github.com/Ineedanamehere/SAE-SSV. |
| title | SAE-SSV: Supervised Steering in Sparse Representation Spaces for Reliable Control of Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2505.16188 |