English
Related papers

Related papers: Steered LLM Activations are Non-Surjective

200 papers

Large language models possess strong chemical reasoning capabilities, making them effective molecular editors. However, property-relevant information is implicitly entangled across their dense hidden states, providing no explicit handle for…

Machine Learning · Computer Science 2026-05-12 Mingxu Zhang , Yuhan Li , Lujundong Li , Dazhong Shen , Hui Xiong , Ying Sun

Robust alignment guardrails for large language models (LLMs) are becoming increasingly important with their widespread application. In contrast to previous studies, we demonstrate that inference-time activation interventions can bypass…

Computation and Language · Computer Science 2025-08-26 Paul Darm , Annalisa Riccardi

Despite impressive breadth, LLMs still rely on explicit reasoning instructions or static, one-fits-all steering methods, leaving a gap for adaptive, instruction-free reasoning amplification. We present Prototype-Based Dynamic Steering…

Computation and Language · Computer Science 2025-10-08 Ceyhun Efe Kayan , Li Zhang

As chain-of-thought (CoT) has become central to scaling reasoning capabilities in large language models (LLMs), it has also emerged as a promising tool for interpretability, suggesting the opportunity to understand model decisions through…

Artificial Intelligence · Computer Science 2026-03-03 Kyle Cox , Darius Kianersi , Adrià Garriga-Alonso

We theoretically predict, and experimentally verify with entangled photons, that outcome communication is not enough for hidden-state models to reproduce quantum steering. Hidden-state models with outcome communication correspond, in turn,…

Quantum Physics · Physics 2018-04-11 R. V. Nery , M. M. Taddei , R. Chaves , L. Aolita

Prompt engineering and finetuning aim to maximize language model performance on a given metric (like toxicity reduction). However, these methods do not fully elicit a model's capabilities. To reduce this gap, we introduce activation…

Computation and Language · Computer Science 2024-10-11 Alexander Matt Turner , Lisa Thiergart , Gavin Leech , David Udell , Juan J. Vazquez , Ulisse Mini , Monte MacDiarmid

Latent steering exploits internal representations of Large Language Models (LLMs) to guide generation, yet interventions on dense states can entangle distinct semantic features. In this paper, we investigate attention query activations as a…

Machine Learning · Computer Science 2026-05-25 Sumanta Bhattacharyya , Pedram Rooshenas

Large language models frequently produce errors in reasoning tasks despite possessing the underlying knowledge required for correct reasoning. One possible approach to improve reasoning consistency is through activation steering. However,…

Machine Learning · Computer Science 2026-05-22 Ian Li , Kapilesh Guruprasad , Raunak Sengupta , Ninad Satish , Loris D'Antoni , Rose Yu

Safety alignment is indispensable for Large Language Models (LLMs) to defend threats from malicious instructions. However, recent researches reveal safety-aligned LLMs prone to reject benign queries due to the exaggerated safety issue,…

Artificial Intelligence · Computer Science 2024-12-18 Zouying Cao , Yifei Yang , Hai Zhao

Direct Prompt Injection (DPI) attacks pose a critical security threat to Large Language Models (LLMs) due to their low barrier of execution and high potential damage. To address the impracticality of existing white-box/gray-box methods and…

Artificial Intelligence · Computer Science 2025-09-10 Minghui Li , Hao Zhang , Yechao Zhang , Wei Wan , Shengshan Hu , pei Xiaobing , Jing Wang

Language models often exhibit undesirable behavior, e.g., generating toxic or gender-biased text. In the case of neural language models, an encoding of the undesirable behavior is often present in the model's representations. Thus, one…

Machine Learning · Computer Science 2025-06-05 Shashwat Singh , Shauli Ravfogel , Jonathan Herzig , Roee Aharoni , Ryan Cotterell , Ponnurangam Kumaraguru

Latent Action Models (LAMs) have rapidly gained traction as an important component in the pre-training pipelines of leading Vision-Language-Action models. However, they fail when observations contain action-correlated distractors, often…

Large language models (LLMs) are known to produce varying responses depending on prompt phrasing, indicating that subtle guidance in phrasing can steer their answers. However, the impact of this framing bias on LLM-based evaluation, where…

Computation and Language · Computer Science 2026-01-21 Yerin Hwang , Dongryeol Lee , Taegwan Kang , Minwoo Lee , Kyomin Jung

Activation steering methods were shown to be effective in conditioning language model generation by additively intervening over models' intermediate representations. However, the evaluation of these techniques has so far been limited to…

Computation and Language · Computer Science 2024-12-02 Daniel Scalena , Gabriele Sarti , Malvina Nissim

As the development and application of Large Language Models (LLMs) continue to advance rapidly, enhancing their trustworthiness and aligning them with human preferences has become a critical area of research. Traditional methods rely…

Computation and Language · Computer Science 2024-11-06 Yuxin Xiao , Chaoqun Wan , Yonggang Zhang , Wenxiao Wang , Binbin Lin , Xiaofei He , Xu Shen , Jieping Ye

Large reasoning models (LRMs) generate chain-of-thought (CoT) traces before producing final outputs, introducing a dynamic internal state that may complicate control mechanisms such as refusal. Unlike instruction-tuned LLMs, where refusal…

Artificial Intelligence · Computer Science 2026-05-27 Kia-Jüng Yang , Dominik Meier , Jiachen Zhao , Terry Ruas , Bela Gipp

Effective and reliable control over large language model (LLM) behavior is a significant challenge. While activation steering methods, which add steering vectors to a model's hidden states, are a promising approach, existing techniques…

Machine Learning · Computer Science 2025-04-03 Samuel Soo , Chen Guang , Wesley Teng , Chandrasekaran Balaganesh , Tan Guoxian , Yan Ming

Subliminal learning describes a student language model inheriting a behavioral bias by fine-tuning on seemingly innocuous data generated by a biased teacher model. Prior work has begun to characterize this phenomenon but leaves open…

Computation and Language · Computer Science 2026-04-29 George Morgulis , John Hewitt

Reasoning-focused LLMs sometimes alter their behavior when they detect that they are being evaluated, which can lead them to optimize for test-passing performance or to comply more readily with harmful prompts if real-world consequences…

Computation and Language · Computer Science 2025-10-29 Sahar Abdelnabi , Ahmed Salem

Large Language Models (LLMs) achieve remarkable performance through pretraining on extensive data. This enables efficient adaptation to diverse downstream tasks. However, the lack of interpretability in their underlying mechanisms limits…

Computation and Language · Computer Science 2025-06-03 Xintong Wang , Jingheng Pan , Liang Ding , Longyue Wang , Longqin Jiang , Xingshan Li , Chris Biemann