paper-with-me

Papers

Steering Vector Fields for Context-Aware Inference-Time Control in Large Language Models

2026-02-02 · Jiaqian Li, Yanshu Li, Kuan-Hao Huang arxiv

Steering vectors (SVs) offer a lightweight way to control large language models (LLMs) at inference time by shifting hidden activations, providing a practical middle ground between prompting and fine-tuning. Yet SVs can be unreliable in practice. Some concepts are unsteerable, and even when steering helps on average it can backfire for a non-trivial fraction of inputs. Reliability also degrades in long-form generation and multi-attribute steering. We take a geometric view of these failures. A static SV applies the same update vector everywhere in representation space, implicitly assuming that the concept-improving direction is constant across contexts. When the locally effective direction varies with the current activation, a single global vector can become misaligned, which yields weak or reversed effects. Guided by this perspective, we propose Steering Vector Fields (SVF), which learns a differentiable concept scoring function whose local gradient defines the steering direction at each activation, making interventions explicitly context-dependent. This formulation supports coordinated multi-layer interventions in a shared, aligned concept space, and enables efficient long-form and multi-attribute control within a unified framework. Across multiple LLMs and steering tasks, SVF delivers stronger and more reliable control, improving the practicality of inference-time steering.

📄 PDF Abstract BibTeX arXiv:2602.01654

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

SALSA: Speech Aware LLM Adaptation via Learned Steering Activation Vectors

2026-05-30 · Yekaterina Yegorova, Argyrios Gerogiannis, Haolong Zheng, Julia Hockenmaier 외 arxiv

Speech-aware large language models often generalize poorly to out-of-domain settings. We propose SALSA (Speech-Aware LLM Adaptation via Learned Steering Activations), a lightweight adaptation method that learns layer-wis…

Dynamic Multimodal Activation Steering for Hallucination Mitigation in Large Vision-Language Models

2026-02-25 · Jianghao Yin, Qin Chen, Kedi Chen, Jie Zhou 외 arxiv

Large Vision-Language Models (LVLMs) exhibit outstanding performance on vision-language tasks but struggle with hallucination problems. Through in-depth analysis of LVLM activation patterns, we reveal two key findings: 1…

Semantic Similarity

DFKI-MLT at SemEval-2026 TASK 7: Steering Multilingual Models Towards Cultural Knowledge

2026-05-21 · Yusser Al Ghussin, Daniil Gurgurov, Yasser Hamidullah, Josef van Genabith 외 arxiv

Large language models (LLMs) are increasingly used across diverse linguistic and cultural contexts, yet their cultural knowledge remains uneven across regions and languages. We present the DFKI-MLT system for SemEval-202…

On the Limitations of Steering in Language Model Alignment

2025-05-02 · Chebrolu Niranjan, Kokil Jaidka, Gerard Christopher Yeo

Steering vectors are a promising approach to aligning language model behavior at inference time. In this paper, we propose a framework to assess the limitations of steering vectors as alignment mechanisms. Using a framew…

Language ModelingLanguage Modelling

Steering Evaluation-Aware Language Models to Act Like They Are Deployed

2025-10-23 · Tim Tian Hua, Andrew Qin, Samuel Marks, Neel Nanda arxiv

Large language models (LLMs) can sometimes detect when they are being evaluated and adjust their behavior to appear more aligned, compromising the reliability of safety evaluations. In this paper, we show that adding a s…