# HyperSteer: Activation Steering at Scale with Hypernetworks

Source: [HyperSteer: Activation Steering at Scale with Hypernetworks](<https://www.sota2.com/research/paper/68fa79111e209131cea7bf3d>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T04:52:00.639Z

Authors: Jiuding Sun, Sidharth Baskaran, Zhengxuan Wu, Michael Sklar, Christopher Potts, Atticus Geiger

Published: 2025-06-03T18:32:01.000Z

## Abstract

Steering language models (LMs) by modifying internal activations is a popular approach for controlling text generation. Unsupervised dictionary learning methods, e.g., sparse autoencoders, can be scaled to produce many steering vectors, but lack guarantees on the individual efficacy of each vector and control over the coverage of relevant steering tasks. In contrast, supervised methods for constructing steering vectors are targeted and effective, but require more data collection and training for each additional steering vector produced. In this work, we introduce HyperSteer, a family of hypernetwork-based architectures which are trained end-to-end to generate steering vectors conditioned on the natural language steering prompts and the internals of the steered LM. In our evaluations, we show that scaling HyperSteer with thousands of steering prompts exceeds the performance of state-of-the-art activation steering methods, even on steering prompts never seen during training. Moreover, HyperSteer performs on par with steering-via-prompting.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Concept Steering on AxBench (Held-in)](<https://www.sota2.com/research/sota/concept-steering-on-axbench-held-in.md>): Concept Steering; AxBench (Held-in); HMean: 1.091.

- [Activation Steering on AxBench Gemma-2-9B layer 20](<https://www.sota2.com/research/sota/activation-steering-on-axbench-gemma-2-9b-layer-20.md>): Activation Steering; AxBench Gemma-2-9B layer 20; Steering Score: 1.091.

- [Concept Steering on AxBench (Held-out)](<https://www.sota2.com/research/sota/concept-steering-on-axbench-held-out.md>): Concept Steering; AxBench (Held-out); HMean: 0.934.

- [Activation Steering on AxBench Gemma-2-2B layer 20](<https://www.sota2.com/research/sota/activation-steering-on-axbench-gemma-2-2b-layer-20.md>): Activation Steering; AxBench Gemma-2-2B layer 20; Steering Score: 0.742.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2506.03292>)

[PDF](<https://arxiv.org/pdf/2506.03292.pdf>)