# The Effectiveness of Style Vectors for Steering Large Language Models: A Human Evaluation

arXiv: https://arxiv.org/abs/2601.21505 · HTML: https://mlsanity.com/paper/2601.21505/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Activation steering methods, one of 13 directions in Mechanistic interpretability.
- Builds on 16 of its 57 references that are in the map.
- First posted on arXiv 29 Jan 2026.

## Placed in

- Mechanistic interpretability / [Activation steering methods](https://mlsanity.com/mechinterp/activation-steering-methods/)

## Builds on (most cited first, 16 of 16)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Prefix-Tuning: Optimizing Continuous Prompts for Generation](https://mlsanity.com/paper/2101.00190/) (2021)
- [Steering Llama 2 via Contrastive Activation Addition](https://mlsanity.com/paper/2312.06681/) (2023)
- [Plug and Play Language Models: A Simple Approach to Controlled Text Generation](https://mlsanity.com/paper/1912.02164/) (2019)
- [CTRL: A Conditional Transformer Language Model for Controllable Generation](https://mlsanity.com/paper/1909.05858/) (2019)
- [GoEmotions: A Dataset of Fine-Grained Emotions](https://mlsanity.com/paper/2005.00547/) (2020)
- [FUDGE: Controlled Text Generation With Future Discriminators](https://mlsanity.com/paper/2104.05218/) (2021)
- [Extracting Latent Steering Vectors from Pretrained Language Models](https://mlsanity.com/paper/2205.05124/) (2022)
- [Large Pre-trained Language Models Contain Human-like Biases of What is Right and Wrong to Do](https://mlsanity.com/paper/2103.11790/) (2021)
- [Semantics-Adaptive Activation Intervention for LLMs via Dynamic Steering Vectors](https://mlsanity.com/paper/2410.12299/) (2024)
- [Style Vectors for Steering Generative Large Language Model](https://mlsanity.com/paper/2402.01618/) (2024)
- [Hooks in the Headline: Learning to Generate Headlines with Controlled Styles](https://mlsanity.com/paper/2004.01980/) (2020)
- [Extending Activation Steering to Broad Skills and Multiple Behaviours](https://mlsanity.com/paper/2403.05767/) (2024)
- [Steering off Course: Reliability Challenges in Steering Language Models](https://mlsanity.com/paper/2504.04635/) (2025)
- [Structuring Latent Spaces for Stylized Response Generation](https://mlsanity.com/paper/1909.05361/) (2019)

## Built on by (most cited first, 4 of 4)

- [Deployable Per-Instance Multi-Layer Activation Steering for Large Language Models](https://mlsanity.com/paper/2608.08829/) (2026)
- [Read-Best Is Not Steer-Best: A Probing--Steering Layer Dissociation in Omni-Modal Large Language Models](https://mlsanity.com/paper/2609.22135/) (2026)
- [Controllable Affective Generation via Latent Vector Steering](https://mlsanity.com/paper/2608.25569/) (2026)
- [First-Order Steering: Translating Weight Adaptation into Activation Steering](https://mlsanity.com/paper/2610.04283/) (2026)

## Abstract (arXiv)

Controlling the behavior of large language models (LLMs) at inference time is essential for aligning outputs with human abilities and safety requirements. Activation steering provides a lightweight alternative to prompt engineering and fine-tuning by directly modifying internal activations to guide generation. This research advances the literature in three significant directions. First, while previous work demonstrated the technical feasibility of steering emotional tone using automated classifiers, this paper presents the first human evaluation of activation steering concerning the emotional tone of LLM outputs, collecting over 7,000 crowd-sourced ratings from 190 participants via Prolific (n=190). These ratings assess both perceived emotional intensity and overall text quality. Second, we find strong alignment between human and model-based quality ratings (mean r=0.776, range 0.157--0.985), indicating automatic scoring can proxy perceived quality. Moderate steering strengths (λ ≈ 0.15) reliably amplify target emotions while preserving comprehensibility, with the strongest effects for disgust (ηp2 = 0.616) and fear (ηp2 = 0.540), and minimal effects for surprise (ηp2 = 0.042). Finally, upgrading from Alpaca to LlaMA-3 yielded more consistent steering with significant effects across emotions and strengths (all p < 0.001). Inter-rater reliability was high (ICC = 0.71--0.87), underscoring the robustness of the findings. These findings support activation-based control as a scalable method for steering LLM behavior across affective dimensions.