Activation Steering

Follow this topic Ask about this topic

Activation steering studies how edits to a model’s hidden activations can control its behavior at inference time, enabling changes without retraining the full model. Common approaches add or remove contrastive steering vectors, target sparse autoencoder features, neurons or attention heads, and use conditional or feedback-based controllers to adapt interventions to the input or generation state. Researchers apply these methods to language tasks such as reasoning, refusal, factuality and hallucination mitigation, as well as image and video generation, speech, and embodied agents; evaluations include benchmarks such as GSM8K, TruthfulQA, POPE and LIBERO. A central challenge is that readable or decodable features are not necessarily controllable, and effective edits can vary by context and architecture or cause collateral changes to fluency, utility and safety.

Latest papers

376 papers77 papers in the last four weeks, up 93% on the four weeks before. 0.5% of all new papers.

Manifold-Guided Attention Steering
May 20, 2026Ian Li, Kapilesh Guruprasad, Raunak Sengupta +3Language Model Steering Self-Attention
University of California, San Diego
Playing Devil's Advocate: Off-the-Shelf Persona Vectors Rival Targeted Steering for Sycophancy
May 20, 2026Ishaan Kelkar, Vikram Kakaria, Nebras Alam +3LLM Sycophancy Language Model Steering
University of Toronto, Toronto, Canada · Princeton University, Princeton, USA · Purdue University, West Lafayette, USA +3
The Hidden Signal of Verifier Strictness: Controlling and Improving Step-Wise Verification via Selective Latent Steering
May 20, 2026Yefan Zhou, Yilun Zhou, Austin Xu +3Language Model Steering Activation Steering
Dartmouth College · Datadog AI Research · Salesforce AI Research
Adaptive Probe-based Steering for Robust LLM Jailbreaking
May 19, 2026Junxi Chen, Junhao Dong, Xiaohua XieLanguage Model Steering Adversarial Attacks on LLMs
School of Computer Science and Engineering, Sun Yat-Sen University, China · Nanyang Technological University, Singapore.
Safety Geometry Collapse in Multimodal LLMs and Adaptive Drift Correction
May 18, 2026Jiahe Guo, Xiangran Guo, Jiaxuan Chen +6Multimodal Large Language Models LLM Safety
Harbin Institute of Technology · Huawei Technologies Co., Ltd
FishBack: Pullback Fisher Geometry for Optimal Activation Steering in Transformers
May 17, 2026Sihan Wang, Jiayi ZhaoInformation Geometry Language Model Steering
South China University of Technology
Contrastive Conceptor Activation Steering (COAST): Unlocking Vision-Language-Action Models through Hidden States
May 16, 2026Miranda Muqing Miao, Subin Kim, Brandon Yang +1VLM Adaptation Vision-Language-Action Models
University of Pennsylvania
Diagnosing and Correcting Concept Omission in Multimodal Diffusion Transformers
May 14, 2026Kanghyun Baek, Jaihyun Lew, Chaehun Shin +2Diffusion Transformer T2I Generation
Interdisciplinary Program in Artificial Intelligence, Seoul National University, Seoul, South Korea · Department of Electrical and Computer Engineering, Seoul National University, Seoul, South Korea · Department of Computer Science & Engineering, Korea University, Seoul, South Korea +1
Few Channels Draw The Whole Picture: Revealing Massive Activations in Diffusion Transformers
May 13, 2026Evelyn Turri, Davide Bucciarelli, Sara Sarto +2Diffusion Transformer Massive Activations
University of Modena and Reggio Emilia, Italy · University of Pisa, Italy
Tracing Persona Vectors Through LLM Pretraining
May 13, 2026Viktor Moskvoretskii, Dominik Glandorf, Jorge Medina Moreira +2Language Model Pretraining LLM Interpretability
EPFL
Inference-Time Machine Unlearning via Gated Activation Redirection
May 12, 2026Vinícius Conte Turani, Otávio Parraga, João Vitor Boer Abitante +7Activation Steering Machine Unlearning
MALTA, Machine Learning Theory and Applications Lab, PUCRS, Porto Alegre, Brazil · Harvard University · Kunumi Institute, Brazil
Inline Critic Steers Image Editing
May 12, 2026Weitai Kang, Xiaohang Zhan, Yizhou Wang +4Text-Guided Image Editing Activation Steering
University of Illinois Chicago · Adobe
OGLS-SD: On-Policy Self-Distillation with Outcome-Guided Logit Steering for LLM Reasoning
May 12, 2026Yuxiao Yang, Xiaoyun Wang, Weitong ZhangLanguage Model Calibration Language Model Distillation
UNC Chapel Hill · NVIDIA
On the Interpretability of Whisper Encodings Using Sparse Autoencoders
May 12, 2026Dan Pluth, Zachary Nicholas Houghton, Yu Zhou +1Transformer Interpretability Mechanistic Interpretability
Vail Systems, Inc. · University of Oregon
Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models
May 12, 2026Boyi Deng, Xu Wang, Yaoning Wang +15LLM Interpretability Sparse Autoencoders
Qwen Team
Compositional Literary Primitives in Instruction-Tuned LLMs: Cross-Architectural SAE Features for Self, Style, and Affect
May 11, 2026Joao Paulo Cavalcante Presa, Savio Salvarino Teles de OliveiraLLM Interpretability Personality Modeling in Language Models
Federal University of Goias (UFG)
Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions
May 11, 2026Diancheng Kang, Zheyuan Liu, Ningshan Ma +3Language Model Steering Self-Attention
1Southern University of Science and Technology · University of Notre Dame · 3Massachusetts Institute of Technology
Cross-Family Universality of Behavioral Axes via Anchor-Projected Representations
May 11, 2026Su-Hyeon Kim, Yo-Sub HanLLM Interpretability Transfer Learning
Department of Artificial Intelligence Yonsei University Seoul, South Korea · Department of Computer Science Yonsei University Seoul, South Korea
Through the Lens of Character: Resolving Modality-Role Interference in Multimodal Role-Playing Agent
May 10, 2026Yihong Tang, Kehai Chen, Xuefeng Bai +1Multimodal Large Language Models Multimodal Agents
Harbin Institute of Technology, Shenzhen, China · Shenzhen Loop Area Institute (SLAI), Shenzhen, China
When Is Rank-1 Steering Cheap? Geometry, Granularity, and Budgeted Search
May 9, 2026John T. Robertson, Jianing Zhu, Haris Vikalo +1Neural Representation Geometry Activation Steering
The University of Texas at Austin
Inference Time Causal Probing in LLMs
May 8, 2026Sadegh Khorasani, Saber Salehkaleybar, Negar Kiyavash +1Causal Interventions in Language Models Activation Steering
School of Computer and Communication Sciences, EPFL„ Lausanne, Switzerland · 2Leiden Institute of Advanced Computer Science (LIACS), Leiden University, , Leiden, The Netherlands · College of Management of Technology, EPFL, Lausanne, Switzerland
HyperTransport: Amortized Conditioning of T2I Generative Models
May 7, 2026Valentino Maiorca, Eleonora Gualdoni, Xavier Suau +3T2I Generation Hypernetworks
1ISTA · 2Apple
Don't Lose Focus: Activation Steering via Key-Orthogonal Projections
May 7, 2026Haoyan Luo, Mateo Espinosa Zarlenga, Mateja JamnikLanguage Model Steering Self-Attention
University of Cambridge · University of Oxford
Towards Steering without Sacrifice: Principled Training of Steering Vectors for Prompt-only Interventions
May 7, 2026Yuntai Bao, Qinfeng Li, Xinyan Yu +6Language Model Steering Causal Interventions in Language Models
Zhejiang University · Innovation and Management Center, School of Software Technology (Ningbo), Zhejiang University · Ant Group
TACT: Mitigating Overthinking and Overacting in Coding Agents via Activation Steering
May 7, 2026Yuan Sui, Yulin Chen, Yibo Li +6AI Coding Agents Long-Horizon LLM Agents
National University of Singapore · Peking University · Meta
Knowing but Not Correcting: Routine Task Requests Suppress Factual Correction in LLMs
May 7, 2026Zixuan Chen, Hao Lin, Zizhe Chen +6LLM Self-Correction LLM Reliability
1Shanghai Jiao Tong University · 2The Chinese University of Hong Kong · 3Ant Group
Beyond Steering Vector: Flow-based Activation Steering for Inference-Time Intervention
May 7, 2026Zehao Jin, Ruixuan Deng, Junran Wang +2Language Model Steering Activation Steering
Georgia Institute of Technology
Chain of Risk: Safety Failures in Large Reasoning Models and Mitigation via Adaptive Multi-Principle Steering
May 7, 2026Xiaomin Li, Jianheng Hou, Zheyuan Deng +6LLM Safety LLM Safety Evaluation
Harvard University · University of Southern California · Brown University +6
Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior
May 6, 2026Daniel Wurgaft, Can Rager, Matthew Kowal +13Neural Representation Geometry Manifold Learning
Stanford University · University College London · Northeastern University +2
Conceptors for Semantic Steering
May 6, 2026Ilias Triantafyllopoulos, Young-Min Cho, Ren Tao +6Language Model Steering Representation Steering
1New York University · University of Pennsylvania

May 20, 2026J/K move · Enter open · S save

Manifold-Guided Attention Steering

Ian Li, Kapilesh Guruprasad, Raunak Sengupta, Ninad Satish+2

University of California, San Diego

Language Model Steering Self-Attention Activation Steering LLM Reasoning

Open paper AskPDF arXiv

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, existing activation steering approaches apply fixed, pre-computed correction vectors, ignoring where the model currently sits along its generation trajectory; the result is indiscriminate perturbation that disrupts already-correct steps as freely as erroneous ones. We propose Manifold-Guided Attention Steering (MAGS), a trajectory-aware inference-time intervention grounded in a geometric observation: the output activations of specific attention heads diverge from a low-dimensional correctness manifold at the point of error, and this deviation compounds through subsequent steps. For each identified attention head, we learn a low-dimensional subspace from contrastive pairs of correct and incorrect traces that capture the directions along which error behavior deviates from correct behavior. During inference, we monitor each head's proximity to this manifold and apply a targeted projection correction when deviation exceeds a learned threshold, steering the attention output back toward the correct subspace before the error propagates. MAGS consistently outperforms both unsteered baselines and static steering approaches across benchmarks spanning mathematical reasoning (MATH-500, GSM8K), code generation (HumanEval, MBPP), and molecular generation (SMILES), suggesting that correctness manifolds are a general feature of LLM attention geometry.