# Beyond Steering Vector: Flow-based Activation Steering for Inference-Time Intervention

Source: [Beyond Steering Vector: Flow-based Activation Steering for Inference-Time Intervention](<https://www.sota2.com/research/paper/69fd6e455497b3b670c8960e>)

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

Data updated: 2026-05-09T05:00:04.796Z

Authors: Zehao Jin, Ruixuan Deng, Junran Wang, Xinjie Shen, Chao Zhang

Published: 2026-05-07T09:03:16.000Z

## Abstract

Activation steering has emerged as a promising alternative for controlling language-model behavior at inference time by modifying intermediate representations while keeping model parameters frozen. However, large-scale evaluations such as AxBench show that existing steering methods are often outperformed by simple in-context prompting and generalize poorly to unseen concepts. We hypothesize that these limitations arise from unvalidated simplifying assumptions shared across prior methods, which typically restrict steering interventions to fixed, single-step, position-invariant transforms. We propose FLAS (Flow-based Activation Steering), which learns a general, concept-conditioned velocity field $v\_t(h,t,c)$ that transports unsteered activations to steered ones without relying on these assumptions. On AxBench, FLAS is the first learned method to consistently outperform prompting, reaching held-out harmonic means of $1.015$ on Gemma-2-2B-IT and $1.113$ on Gemma-2-9B-IT without per-concept tuning. Analysis of the learned flow shows curved, multi-step, token-varying trajectories, which suggests that previous hypotheses on activation space geometry might be incomplete.

## 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: 1.185.

- [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: 1.113.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.05892>)

[PDF](<https://arxiv.org/pdf/2605.05892.pdf>)