Steer Like the LLM: Activation Steering that Mimics Prompting

Geert Heyman, Frederik Vandeputte
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:42968-42997, 2026.

Abstract

Large language models can be steered at inference time through prompting or activation interventions, but activation steering methods often underperform compared to prompt-based approaches. We investigate whether activation steering can be improved by learning to mimic the interventions that prompt steering triggers within the model. To this end, we introduce Prompt Steering Replacement (PSR) models, a new family of activation steering methods that distill prompt steering behavior into interpretable interventions on model activations. A PSR is an activation steering method that estimates position-specific steering coefficients and is trained to imitate prompt-based interventions. Experiments on persona steering and instruction following across multiple language models demonstrate that PSR models consistently outperform constant-coefficient interventions that are frequently used in the literature and achieve performance close to or exceeding prompt steering while maintaining interpretability.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-heyman26a, title = {Steer Like the {LLM}: Activation Steering that Mimics Prompting}, author = {Heyman, Geert and Vandeputte, Frederik}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {42968--42997}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/heyman26a/heyman26a.pdf}, url = {https://proceedings.mlr.press/v306/heyman26a.html}, abstract = {Large language models can be steered at inference time through prompting or activation interventions, but activation steering methods often underperform compared to prompt-based approaches. We investigate whether activation steering can be improved by learning to mimic the interventions that prompt steering triggers within the model. To this end, we introduce Prompt Steering Replacement (PSR) models, a new family of activation steering methods that distill prompt steering behavior into interpretable interventions on model activations. A PSR is an activation steering method that estimates position-specific steering coefficients and is trained to imitate prompt-based interventions. Experiments on persona steering and instruction following across multiple language models demonstrate that PSR models consistently outperform constant-coefficient interventions that are frequently used in the literature and achieve performance close to or exceeding prompt steering while maintaining interpretability.} }
Endnote
%0 Conference Paper %T Steer Like the LLM: Activation Steering that Mimics Prompting %A Geert Heyman %A Frederik Vandeputte %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-heyman26a %I PMLR %P 42968--42997 %U https://proceedings.mlr.press/v306/heyman26a.html %V 306 %X Large language models can be steered at inference time through prompting or activation interventions, but activation steering methods often underperform compared to prompt-based approaches. We investigate whether activation steering can be improved by learning to mimic the interventions that prompt steering triggers within the model. To this end, we introduce Prompt Steering Replacement (PSR) models, a new family of activation steering methods that distill prompt steering behavior into interpretable interventions on model activations. A PSR is an activation steering method that estimates position-specific steering coefficients and is trained to imitate prompt-based interventions. Experiments on persona steering and instruction following across multiple language models demonstrate that PSR models consistently outperform constant-coefficient interventions that are frequently used in the literature and achieve performance close to or exceeding prompt steering while maintaining interpretability.
APA
Heyman, G. & Vandeputte, F.. (2026). Steer Like the LLM: Activation Steering that Mimics Prompting. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:42968-42997 Available from https://proceedings.mlr.press/v306/heyman26a.html.

Related Material