cs.HCOct 3, 2026

DimSteer: Steering LLM Authoring with Automatically Discovered Stylistic Controls

Authors: Ajit Mallavarapu, Ziwei Gu

Organizations: Harvard University, USA

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Large language model writing interfaces often make users steer outputs by repeatedly articulating desired changes in natural language. Yet writers may recognize useful stylistic directions only after seeing alternatives, making revision recall-heavy. We present DimSteer, an authoring interface that samples prompt-local completions, discovers high-variance activation-space axes of variation, labels them, and exposes them as sliders with pole previews, diff comparison, and reset controls. Users can manipulate discovered dimensions, reducing the need to reformulate prompts for each stylistic adjustment. In a within-subjects study with 16 participants against a matched prompt-only baseline, DimSteer reduced mental demand, effort, and frustration while preserving comparable perceived success. Participants valued the surfaced dimensions, yet 15 of 16 disagreed that they would have thought to request the same changes in a prompt. Results suggest prompt-local controls can shift LLM authoring from recall-based prompting toward recognition-based exploration and direct manipulation, while preserving prompting for open-ended edits.

Figures & tables

Figure 1. DimSteer shifts LLM interaction from recall-based prompting to direct manipulation. (1) Standard chat interfaces require users to articulate the exact changes needed to steer text. (2) DimSteer instead surfaces automatically discovered dimensions of generation variation as continuous controls, allowing users to explore variation through recognition and direct manipulation. (3) When discovered controls are insufficient, users can define custom dimensions to extend the machine-proposed control space. A three-stage overview of DimSteer showing a transition from recall-based prompting, to automatically discovered semantic sliders, to user-defined repair and extension of the control space.
Figure 2. Interface A: Prompt-Only Baseline. Participants revised drafts using standard natural-language instructions.
Figure 3. The DimSteer Interface. (A) Each discovered slider card includes (A1) current steering coefficient, (A2) PCA variance tag, and (A3) directional pole text previews. (B) Reset returns steering to the baseline. (C) Custom-dimension field lets users add missing controls. (D) Diff toggle reveals changes from the baseline draft.
Figure 4. DimSteer implementation pipeline. The system samples prompt-local completions, extracts layer-21 activations, identifies PCA axes, labels and orients them with an LLM judge, and applies the resulting vectors through slider-based activation steering.
PromptTask promptSurfaced dimensions
P1Describe the chaotic energy of a packed night market in a bustling city.Vivid, Sensational, Atmosphere
P2Describe a sudden thunderstorm hitting a quiet, open field during a hot afternoon.Vividness, Sensory, Dissonance
P3Write a brief, encouraging announcement message for a team channel celebrating a successful project launch.Formality, Enthusiasm, Tone
P4Write a brief, urgent Slack message asking teammates to hop onto a bridge call to fix a sudden production crash.Urgency, Formality, Directness
Table 1. Shared study prompts and surfaced dimensions.
MeasureCPROMPTC_{PROMPT}CSTEERC_{STEER}
Mean task duration5.72 min3.22 min
Median task duration5.70 min3.20 min
Task duration range5.3–6.2 min3.0–3.5 min
Median revision actions3.5 reprompts4.5 slider changes
Revision action range3–42–7
Diff view used11/1613/16 participants
Table 2. Descriptive interaction-log summaries across participants and tasks.
Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

PromptPCA-discovered controlsGPT5.6-luna-generated controls
P1Vivid; Sensational; AtmosphereImmersiveness; Sensory Intensity; Pacing
P2Vividness; Sensory; DissonanceDramatic Contrast; Sensory Detail; Atmospheric Tension
P3Formality; Enthusiasm; ToneWarmth; Enthusiasm; Formality
P4Urgency; Formality; DirectnessDirectness; Reassurance; Urgency
Appendix
Table 3. Control sets shown in the supplementary control-source evaluation.
More in this topic

Language Model Steering

Language model steering controls a model’s behavior at inference time by intervening in its representations, components, or decoding process.

See all 279 papers

Explore similar work

CardsList
May 5, 2026

Steer Like the LLM: Activation Steering that Mimics Prompting

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 propose a framework that formulates prompt steering as a form of activation steering and investigates whether distilling successful prompt steering behavior into simpler, interpretable models can close this gap. Our analysis reveals that popular activation steering methods are not faithful to the mechanics of prompt steering, which applies strong interventions on some tokens while barely affecting others. Based on these insights, we introduce Prompt Steering Replacement (PSR) models that estimate token-specific steering coefficients from the activations themselves and are trained to imitate prompt-based interventions. Experiments on three steering benchmarks across multiple language models show that PSR models outperform existing activation steering methods, especially when controlling for high-coherence completions, and also compare favorably to prompting on AxBench and persona steering.

Geert Heyman, Frederik Vandeputte

Nokia Bell Labs, Belgium.

Language Model SteeringLLM PromptingCausal Interventions in Language ModelsActivation Steering
AskPDF
Apr 13, 2026

From Words to Widgets for Controllable LLM Generation

Natural language remains the predominant way people interact with large language models (LLMs). However, users often struggle to precisely express and control subjective preferences (e.g., tone, style, and emphasis) through prompting. We propose Malleable Prompting, a new interactive prompting technique for controllable LLM generation. It reifies preference expressions in natural language prompts into GUI widgets (e.g., sliders, dropdowns, and toggles) that users can directly configure to steer generation, while visualizing each control's influence on the output to support attribution and comparison across iterations. To enable this interaction, we introduce an LLM decoding algorithm that modulates the token probability distribution during generation based on preference expressions and their widget values. Through a user study, we show that Malleable Prompting helps participants achieve target preferences more precisely and is perceived as more controllable and transparent than natural language prompting alone.

Chao Zhang, Yiren Liu, Lunyiu Nie +3

Cornell University Ithaca, NY, USA · University of Illinois Urbana-Champaign Champaign, IL, USA · The University of Texas at Austin Austin, TX, USA +1

LLM PromptingControllable Text GenerationLanguage Model Steering
AskPDF
Sep 4, 2026

Can Activation Steering Capture Multidimensional Authorship Style?

Activation steering has shown promise for controlling LLM generation along well-defined attributes, but it remains unclear whether it can handle the multidimensional and hard-to-define nature of authorship style. We ask whether structured contrastive prompting along rhetorically-motivated dimensions can construct rich style representations directly in activation space, bypassing the need for natural language style descriptors or dedicated training. We find that the resulting directions share a common authorship backbone while conflicting on aspect-specific residuals that carry genuine stylistic signal, explaining why naive aggregation fails. We operationalize this in Aspect-Aware Activation Steering (A3S), a training-free framework that merges per-aspect contrastive directions with interference-aware aggregation and tunes steering strength per instance. A3S improves authorship style transfer where it is genuinely multi-aspect, outperforms a trained baseline in preference evaluations on out-of-domain benchmarks, and keeps target-exemplar overlap consistently low.

Hieu Tran, Calvin Bao, Marine Carpuat

University of Maryland, College Park

Language Model SteeringActivation Steering
AskPDF