arXiv Computation and Language August 25, 2026 By Zhiwei Gao, Shaowen Peng, Shoko Wakamiya, Eiji Aramaki

What Does Activation Steering Control? Attribution Across Answer Encodings and Output-Sensitive Subspaces

Read the original on arXiv Computation and Language →

The paper investigates what aspects of language model behavior are controlled by activation steering. By introducing Cross‑Encoding Steering Evaluation, the authors show that steering effects often follow the extraction index of answer identifiers rather than the semantic content of the answers, especially at deeper layers. They also find that a low‑rank output‑sensitive component captures most of this effect, and that different datasets (NormBank, MNLI, SC101) exhibit varying preferences for extraction‑index versus semantic‑label following.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Computation and Language.

Hugging Face Trending Papers

Aug 24

What Does Activation Steering Control? Attribution Across Answer Encodings and Output-Sensitive Subspaces

Activation steering is often evaluated under the answer encoding used to construct the direction. A reported gain may reflect the intended judgment or compatibility with answer identifiers seen during...

arXiv AI

Jun 11

The Structural Attention Tax: How Retrieval Format Hijacks In-Context Learning Independent of Content

arXiv:2606. 11198v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) systems inject external knowledge to improve LLM outputs, yet the format of injected content -- distinct from its semantic relevance -- can independently distort the model's attention distribution.

By Yuqi Zhang, Di Zhang

llms rag benchmarks safety

arXiv Machine Learning

Aug 5

Inverted Detection and Control in Steering Vectors

arXiv:2608. 02957v1 Announce Type: new Abstract: Steering vectors (SVs) are widely used to influence the expression of concepts (e.

By Max Torop, Aria Masoomi, Jennifer Dy

llms

arXiv AI

Jun 9

Activation Steering Induces Emergent Misalignment: A More Comprehensive Evaluation

arXiv:2606. 08682v1 Announce Type: cross Abstract: Activation steering has emerged as a popular inference-time technique for modulating the behavior of large language models (LLMs).

By Qi Cao, Jian Lou, Meiting Liu, Wenjie Feng, Dan Li, See-Kiong Ng, Anh Tuan Luu

llms fine-tuning safety

arXiv AI

Jul 13

GrAInS: Gradient-based Attribution for Inference-Time Steering of LLMs and VLMs

arXiv:2507. 18043v2 Announce Type: replace-cross Abstract: Inference-time steering methods offer a lightweight alternative to fine-tuning large language models (LLMs) and vision-language models (VLMs) by modifying internal activations at test time without updating model weights.

By Duy Nguyen, Archiki Prasad, Elias Stengel-Eskin, Mohit Bansal

llms fine-tuning multimodal safety

arXiv AI

Jul 7

Detecting Hallucinations in Retrieval-Augmented Generation through Grounding-Aware Sensitivity by Perturbation (GASP)

arXiv:2607. 04223v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) reduces but does not eliminate hallucination, and existing detectors return a single answer-level score that does not indicate which sentence is unsupported, or why.

By Mohamed Aly Bouke

rag nlp benchmarks safety

One email a morning, machine-written

Related stories

What Does Activation Steering Control? Attribution Across Answer Encodings and Output-Sensitive Subspaces

The Structural Attention Tax: How Retrieval Format Hijacks In-Context Learning Independent of Content

Inverted Detection and Control in Steering Vectors

Activation Steering Induces Emergent Misalignment: A More Comprehensive Evaluation

GrAInS: Gradient-based Attribution for Inference-Time Steering of LLMs and VLMs

Detecting Hallucinations in Retrieval-Augmented Generation through Grounding-Aware Sensitivity by Perturbation (GASP)