Steering Geometry: Validating Human Value Geometry in LLM Steering Space

Hugging Face Daily Papers Papers
activation-steering human-values latent-geometry llms steering-vectors alignment benchmark

Summary

This paper introduces a benchmark to validate that distribution-driven methods for activation steering in LLMs recover human value topologies aligned with Schwartz's theory, improving with model size but declining post-instruction tuning.

As large language models (LLMs) are increasingly deployed in alignment-sensitive contexts, activation steering has emerged as a lightweight, inference-time alternative to fine-tuning methods (e.g., RLHF, DPO) for behavioral control. However, existing work typically validates steering on isolated behaviors, leaving it unclear whether steering vectors encode coherent semantic structure or merely exploit behavior-specific shortcuts. We investigate whether the latent geometry of LLM steering vectors reflects theory-specified structure in human values and morality. Using Schwartz's Theory of Basic Human Values as our primary fine-grained framework, we introduce a 26K-sample benchmark covering 20 human values and analyze distribution-driven methods (e.g., CAA, SphericalSteer, ODESteer) and behavior-centric approaches (e.g., COLD-Steer, BiPO) across diverse model families and sizes. We find that distribution-driven methods recover human value topologies aligned with theoretical predictions (Spearman ρ up to 0.51, p < 10^{-13}). In contrast, behavior-centric methods achieve comparable steering performance but show little correlation with the expected value geometry. Geometric fidelity improves with model scale but drops after instruction tuning. Finally, better geometric alignment also leads to more human-consistent transfer across values: steering one value correctly lifts compatible values and suppresses opposing ones. Code and data are available at: https://github.com/DeepRCL/Steering_Geometry.
Original Article
View Cached Full Text

Cached at: 09/09/26, 04:28 AM

Paper page - Steering Geometry: Validating Human Value Geometry in LLM Steering Space

Source: https://huggingface.co/papers/2609.06289

Abstract

Activation steering vectors in large language models encode theory-aligned human value geometry when derived via distribution-driven methods, with geometric fidelity scaling with model size but declining after instruction tuning.

As large language models (LLMs) are increasingly deployed in alignment-sensitive contexts,activation steeringhas emerged as a lightweight, inference-time alternative to fine-tuning methods (e.g., RLHF, DPO) for behavioral control. However, existing work typically validates steering on isolated behaviors, leaving it unclear whethersteering vectorsencode coherent semantic structure or merely exploit behavior-specific shortcuts. We investigate whether thelatent geometryof LLMsteering vectorsreflects theory-specified structure in human values and morality. UsingSchwartz’s Theory of Basic Human Valuesas our primary fine-grained framework, we introduce a 26K-sample benchmark covering 20 human values and analyzedistribution-driven methods(e.g.,CAA,SphericalSteer,ODESteer) and behavior-centric approaches (e.g.,COLD-Steer,BiPO) across diverse model families and sizes. We find thatdistribution-driven methodsrecover human value topologies aligned with theoretical predictions (Spearman ρ up to 0.51, p < 10^{-13}). In contrast, behavior-centric methods achieve comparable steering performance but show little correlation with the expected value geometry. Geometric fidelity improves with model scale but drops afterinstruction tuning. Finally, better geometric alignment also leads to more human-consistent transfer across values: steering one value correctly lifts compatible values and suppresses opposing ones. Code and data are available at: https://github.com/DeepRCL/Steering_Geometry.

View arXiv pageView PDFProject pageGitHub1Add to collection

Get this paper in your agent:

hf papers read 2609\.06289

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.06289 in a model README.md to link it from this page.

Datasets citing this paper1

#### DeepRCL/SteeringGeometry Viewer• Updated43 minutes ago • 32k

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.06289 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

When is Your LLM Steerable?

arXiv cs.CL

This paper investigates when activation steering succeeds or fails for LLMs by analyzing early decoding dynamics. The authors introduce ASTEER, a large testbed of steered generations, and train a GBDT classifier to predict steering outcomes from early hidden states, enabling efficient steering strength search.