Conceptors for Semantic Steering

• URL: http://arxiv.org/abs/2605.04980v1
• Date: Wed, 06 May 2026 14:32:29 GMT
• Title: Conceptors for Semantic Steering
• Abstract summary: Activation-based steering provides control of LLM behavior at inference time.<n>The dominant paradigm reduces each concept to a single direction whose geometry is left largely unexamined.<n>We use conceptors: soft projection matrices estimated from activations pooled across both poles of a bipolar concept.
• Score: 14.40996156810136
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Activation-based steering provides control of LLM behavior at inference time, but the dominant paradigm reduces each concept to a single direction whose geometry is left largely unexamined. Rather than selecting a single steering direction, we use conceptors: soft projection matrices estimated from activations pooled across both poles of a bipolar concept, which preserve the concept's full multidimensional subspace. A geometric analysis shows the bipolar subspace strictly subsumes the single-vector baseline. We further show that the conceptor quota provides a parameter-free layer-selection diagnostic, predicting concept separability with Pearson correlations up to r=0.96 across three instruction-tuned models and three semantic dimensions. Beyond selection, conceptors admit a closed-form Boolean algebra (AND, OR, NOT): we evaluate conceptor compositionality on thematically related sub-concepts. Across a systematic five-axis design-space evaluation, conceptors match or outperform additive baselines at layers where concept subspaces are multi-dimensional while producing substantially fewer degenerate outputs. Conceptor steering is a geometrically principled, compositional, and practically safer alternative to single-direction steering from a limited number of contrastive pairs.

Related papers

• Structuring Sparsity: Block-Sparse Featurizers Capture Visual Concept Manifolds [40.072358858272445]
  We show that concepts are often realized as geometric structures in low dimensional regions of activation space.<n>We implement three variants of block-sparse featurizers (BSFs) and, through a minimum-description-length analysis, show that all three describe activations more compactly than direction-based featurizers.
  arXiv  Detail & Related papers  (2026-06-23T23:28:30Z)
• MidSteer: Optimal Affine Framework for Steering Generative Models [53.89402578892719]
  MidSteer is a more general affine framework for concept manipulation that relaxes assumptions and enables directed, minimal-disturbance transformations.<n>We demonstrate that MidSteer performs favorably across a range of tasks, modalities, and architectures.
  arXiv  Detail & Related papers  (2026-04-17T19:23:33Z)
• Sparse Concept Anchoring for Interpretable and Controllable Neural Representations [0.9831489366502301]
  We introduce Sparse Concept Anchoring, a method that biases latent space to position a targeted subset of concepts.<n>The anchored geometry enables two practical interventions: behavioral steering that projects out a concept's latent component at inference, and permanent removal.
  arXiv  Detail & Related papers  (2025-12-13T21:43:17Z)
• A Geometric Unification of Concept Learning with Concept Cones [58.70836885177496]
  Two traditions of interpretability have evolved side by side but seldom spoken to each other: Concept Bottleneck Models (CBMs) and Sparse Autoencoders (SAEs)<n>We show that both paradigms instantiate the same geometric structure.<n>CBMs provide human-defined reference geometries, while SAEs can be evaluated by how well their learned cones approximate or contain those of CBMs.
  arXiv  Detail & Related papers  (2025-12-08T09:51:46Z)
• Toward a Flexible Framework for Linear Representation Hypothesis Using Maximum Likelihood Estimation [3.515066520628763]
  We introduce a new notion of binary concepts as unit vectors in a canonical representation space.<n>Our method, Sum of Activation-base Normalized Difference (SAND), formalizes the use of activation differences modeled as samples from a von Mises-Fisher distribution.
  arXiv  Detail & Related papers  (2025-02-22T23:56:30Z)
• Beyond Scalars: Concept-Based Alignment Analysis in Vision Transformers [10.400355814467401]
  Vision transformers (ViTs) can be trained using various learning paradigms, from fully supervised to self-supervised.<n>We propose a concept-based alignment analysis of representations from four different ViTs.<n>The concept-based alignment analysis of representations from four different ViTs reveals that increased supervision correlates with a reduction in the semantic structure of learned representations.
  arXiv  Detail & Related papers  (2024-12-09T16:33:28Z)
• 360 Layout Estimation via Orthogonal Planes Disentanglement and Multi-view Geometric Consistency Perception [56.84921040837699]
  Existing panoramic layout estimation solutions tend to recover room boundaries from a vertically compressed sequence, yielding imprecise results. We propose an orthogonal plane disentanglement network (termed DOPNet) to distinguish ambiguous semantics. We also present an unsupervised adaptation technique tailored for horizon-depth and ratio representations. Our solution outperforms other SoTA models on both monocular layout estimation and multi-view layout estimation tasks.
  arXiv  Detail & Related papers  (2023-12-26T12:16:03Z)
• Householder Projector for Unsupervised Latent Semantics Discovery [58.92485745195358]
  Householder Projector helps StyleGANs to discover more disentangled and precise semantic attributes without sacrificing image fidelity. We integrate our projector into pre-trained StyleGAN2/StyleGAN3 and evaluate the models on several benchmarks.
  arXiv  Detail & Related papers  (2023-07-16T11:43:04Z)
• Multi-dimensional concept discovery (MCD): A unifying framework with completeness guarantees [1.9465727478912072]
  We propose Multi-dimensional Concept Discovery (MCD) as an extension of previous approaches that fulfills a completeness relation on the level of concepts. We empirically demonstrate the superiority of MCD against more constrained concept definitions.
  arXiv  Detail & Related papers  (2023-01-27T18:53:19Z)
• Concept Activation Regions: A Generalized Framework For Concept-Based Explanations [95.94432031144716]
  Existing methods assume that the examples illustrating a concept are mapped in a fixed direction of the deep neural network's latent space. In this work, we propose allowing concept examples to be scattered across different clusters in the DNN's latent space. This concept activation region (CAR) formalism yields global concept-based explanations and local concept-based feature importance.
  arXiv  Detail & Related papers  (2022-09-22T17:59:03Z)
• Concept Gradient: Concept-based Interpretation Without Linear Assumption [77.96338722483226]
  Concept Activation Vector (CAV) relies on learning a linear relation between some latent representation of a given model and concepts. We proposed Concept Gradient (CG), extending concept-based interpretation beyond linear concept functions. We demonstrated CG outperforms CAV in both toy examples and real world datasets.
  arXiv  Detail & Related papers  (2022-08-31T17:06:46Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.