Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2504.04635 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909568584384512 |
|---|---|
| author | Da Silva, Patrick Queiroz Sethuraman, Hari Rajagopal, Dheeraj Hajishirzi, Hannaneh Kumar, Sachin |
| author_facet | Da Silva, Patrick Queiroz Sethuraman, Hari Rajagopal, Dheeraj Hajishirzi, Hannaneh Kumar, Sachin |
| contents | Steering methods for language models (LMs) have gained traction as lightweight alternatives to fine-tuning, enabling targeted modifications to model activations. However, prior studies primarily report results on a few models, leaving critical gaps in understanding the robustness of these methods. In this work, we systematically examine three prominent steering methods -- DoLa, function vectors, and task vectors. In contrast to the original studies, which evaluated a handful of models, we test up to 36 models belonging to 14 families with sizes ranging from 1.5B to 70B parameters. Our experiments reveal substantial variability in the effectiveness of the steering approaches, with a large number of models showing no improvement and at times degradation in steering performance. Our analysis demonstrate fundamental flaws in the assumptions underlying these methods, challenging their reliability as scalable steering solutions. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_04635 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Steering off Course: Reliability Challenges in Steering Language Models Da Silva, Patrick Queiroz Sethuraman, Hari Rajagopal, Dheeraj Hajishirzi, Hannaneh Kumar, Sachin Computation and Language Steering methods for language models (LMs) have gained traction as lightweight alternatives to fine-tuning, enabling targeted modifications to model activations. However, prior studies primarily report results on a few models, leaving critical gaps in understanding the robustness of these methods. In this work, we systematically examine three prominent steering methods -- DoLa, function vectors, and task vectors. In contrast to the original studies, which evaluated a handful of models, we test up to 36 models belonging to 14 families with sizes ranging from 1.5B to 70B parameters. Our experiments reveal substantial variability in the effectiveness of the steering approaches, with a large number of models showing no improvement and at times degradation in steering performance. Our analysis demonstrate fundamental flaws in the assumptions underlying these methods, challenging their reliability as scalable steering solutions. |
| title | Steering off Course: Reliability Challenges in Steering Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2504.04635 |