Representation Steering

Ask about this topic

Overview

Representation steering studies how changing a model’s hidden activations can alter its outputs at inference time, often without retraining or modifying weights. Common approaches add or project along directions derived from activation differences or probes, edit sparse autoencoder features, and use subspaces or geometry-aware methods to limit unintended changes. Researchers apply these techniques to control refusal, tool use, reasoning, hallucinations, safety, and demographic bias across language, vision, and speech models. A central tension is that a concept being readable in activations does not guarantee that steering it will change the desired behavior cleanly: interventions can harm utility, miss their target, or create collateral effects, so evaluations must measure both behavioral success and side effects.

Latest papers

CardsList

30 papers7 papers in the last four weeks, against 1 the four weeks before. 0.0% of all new papers.

  1. Principled Nonlinear Concept Erasure and Steering

    Oct 23, 2026Idan Simai, Shauli Ravfogel, Uri ShahamConcept ErasurePrivacy-Preserving Representation Learning

    Department of Computer Science, Bar-Ilan University · Center for Data Science, New York University

  2. Retrieval Disagreement as a Signal: Diagnosing and Steering Multilingual Dense Embedders

    Oct 13, 2026Faeze Ghorbanpour, Shaghayegh Kolli, Jana Diesner +1Multilingual IRRepresentation Steering

    School of Computation, Information and Technology, TU Munich · Munich Center for Machine Learning (MCML)

  3. Beyond Visual Enhancement: Adaptive Multi-Context Steering to Mitigate LVLM Hallucinations

    Oct 8, 2026Shuran Ma, JiaLe Li, Yuxin Dong +9VLM Hallucination MitigationLarge Vision-Language Models

    Shanghai Jiao Tong University · Peking University · Beijing University of Chemical Technology +3

  4. Dino Forcing Flow Models: Do not denoise what you can predict

    Oct 8, 2026Arijit Ghosh, Lucas Degeorge, Paul Couairon +3Self-ConditioningFlow-Based Generative Modeling

    ENPC, IP Paris · Ecole Polytechnique, IP Paris · AMIAD +1

  5. Training-Free Instruction TTS Gender Bias Calibration Using Model-Adaptive Steering

    Oct 7, 2026Kuan-Yu Chen, Yi-Cheng Lin, Jeng-Lin Li +1Text-to-SpeechTTS Synthesis

    AI Research Center, Inventec Corporation, Taiwan · Graduate Institute of Communication Engineering, National Taiwan University, Taiwan

  6. Devils in Question Relay: Source-Conditioned Relay Steering to Mitigate Hallucinations in Audio-visual Large Language Models

    Sep 29, 2026Yu Zhang, Pingrui Zhang, Xuefeng Bai +3LLM Hallucination MitigationHallucination in Language Models

    Harbin Institute of Technology, Shenzhen, China

  7. Who Is Left of Whom? Tracing Spatial Evidence and Role Binding in Relative-Position Reasoning

    Sep 28, 2026Yingjin Song, Denis Paperno, Albert GattVLM InterpretabilitySpatial Reasoning

    Utrecht University Utrecht, The Netherlands

  8. A Probe Shift Is Not a Fairness Fix: The Limits of Representation Steering in Speech Models

    Sep 16, 2026Nicolas Bourrel, Abderrahmane Issam, Gerasimos SpanakisASR EvaluationDemographic Bias in ASR

    Department of Advanced Computing Sciences Maastricht University

  9. Tunable Tool-Call Rates in LLM Agents via Representation Steering

    Aug 25, 2026Yuqi Chen, Vincent Siu, Yang Liu +2LLM Tool UseTool Use

    Department of Computer Science and Engineering, UC Santa Cruz · Department of Computer Science, UC Berkeley

  10. NullEdit: Stealthy Image Protection via VLM Condition Redirection

    Aug 11, 2026Weiyao Huang, Liqin Wang, Ziqi Sheng +1VLM RobustnessRepresentation Steering

    School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China

  11. Risky Business: Measuring The Faithfulness-Safety Tension

    Aug 4, 2026Dominik Meier, Luca Joshua Francis, Marco Bernhard Kaiser +3CoT FaithfulnessLanguage Model Safety Evaluation

    University of Göttingen · 2Landeskriminalamt NRW

  12. Architectural Backdoors in Vision-Language Model Supply Chains via Representation Steering

    Jul 28, 2026Maria Rosaria Briglia, Igor Maljkovic, Antonio Emanuele Cinà +3Adversarial Attacks on VLMsRepresentation Engineering

    Sapienza University of Rome, Italy · University of Genoa, Italy · University of Trieste, Italy +1

  13. Controlling Tool Use with Heading-Specific Activation Steering

    Jul 7, 2026Yuqi Chen, Vincent Siu, Yang Liu +2Language Model SteeringLLM Tool Use

    Department of Computer Science and Engineering, UC Santa Cruz · Department of Computer Science, UC Berkeley

  14. Pulling The REINS: Training-Free Safety Alignment of Video Diffusion Models via Representation Steering

    Jun 15, 2026Rohit Kundu, Arindam Dutta, Sarosij Bose +2Diffusion Model AlignmentVideo Diffusion Models

    University of California, Riverside · YouTube (Google)

  15. Predicting Future Behaviors in Reasoning Models Enables Better Steering

    Jun 9, 2026Evgenii Kortukov, Piotr Komorowski, Florian Klein +5Language Model SteeringRepresentation Steering

    1Fraunhofer HHI · 2Northeastern University · 3KAIST

  16. Whisper Hallucination Detection and Mitigation via Hidden Representation Steering and Sparse AutoEncoders

    Jun 5, 2026Georgii Aparin, Vadim Popov, Tasnima Sadekova +1Audio-Language Model Hallucination DetectionSparse Autoencoders

    AI Foundation and Algorithm Lab · National Research University Higher School of Economics

  17. When Should Models Change Their Minds? Contextual Belief Management in Large Language Models

    May 28, 2026Haoming Xu, Weihong Xu, Zongrui Li +6Representation SteeringBelief Updating in LLMs

    ♠Zhejiang University · ♣HomologyAI

  18. Riemannian-Manifold Steering: Geometry-Aware Generative Autoencoders for Label-Free Steering

    May 24, 2026Narmeen Oozeer, Shivam Raval, Philip Quirke +4Language Model SteeringRepresentation Steering

    Martian · Harvard University · Thoughtworks +1

  19. FishBack: Pullback Fisher Geometry for Optimal Activation Steering in Transformers

    May 17, 2026Sihan Wang, Jiayi ZhaoInformation GeometryLanguage Model Steering

    South China University of Technology

  20. SwordBench: Evaluating Orthogonality of Steering Image Representations

    May 10, 2026Vladimir Zaigrajew, Dawid Pludowski, Hubert Baniecki +1Vision Foundation ModelsRepresentation Engineering

    Centre for Credible AI Warsaw University of Technology Warsaw, Poland · University of Warsaw Warsaw, Poland

  21. Don't Lose Focus: Activation Steering via Key-Orthogonal Projections

    May 7, 2026Haoyan Luo, Mateo Espinosa Zarlenga, Mateja JamnikLanguage Model SteeringSelf-Attention

    University of Cambridge · University of Oxford

  22. Conceptors for Semantic Steering

    May 6, 2026Ilias Triantafyllopoulos, Young-Min Cho, Ren Tao +6Language Model SteeringRepresentation Steering

    1New York University · University of Pennsylvania

  23. The Cylindrical Representation Hypothesis for Language Model Steering

    May 3, 2026Lang Gao, Jinghui Zhang, Wei Liu +7Language Model SteeringLLM Interpretability

    Department of Natural Language Processing, Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, UAE · Department of Computer Science, National University of Singapore, Singapore

  24. Knowledge Vector of Logical Reasoning in Large Language Models

    Apr 26, 2026Zixuan Wang, Yuanyuan LeiLLM InterpretabilityRepresentation Engineering

    Department of Computer & Information Science and Engineering University of Florida, Gainesville, FL, United States

  25. Predicting Where Steering Vectors Succeed

    Apr 16, 2026Jayadev BillaLanguage Model SteeringRepresentation Steering

  26. What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal

    Apr 9, 2026Stephen Cheng, Sarah Wiegreffe, Dinesh ManochaLanguage Model SteeringLLM Interpretability

    University of Maryland, College Park

  27. The Information Geometry of Softmax: Probing and Steering

    Feb 17, 2026Kiho Park, Todd Nief, Yo Joong Choe +1Representation Geometry in Language ModelsInformation Geometry

    University of Chicago · INSEAD

  28. Watch the Model Think: On-Policy Extraction of Activation Steering Vectors

    Feb 15, 2026Xuanbo Su, Yingfang Zhang, Hao Luo +3Language Model SteeringRepresentation Steering

    Bairong Inc., Beijing, China · School of Mathematics, Harbin Institute of Technology, Harbin, China

  29. Beyond Forgetting: Representation Misdirection Elicits Controllable Side Behaviors and Capabilities

    Jan 29, 2026Tien Dang, The-Hai Nguyen, Dinh Mai Phuong +5Linear Representation HypothesisControllable Text Generation

    Japan Advanced Institute of Science and Technology · Monash University · Quantum AI and Cyber Security Institute, FPT Corporation +1

  30. Language steering in latent space to mitigate unintended code-switching

    Oct 11, 2025Andrey Goncharov, Nikolai Kondusov, Alexey ZaytsevLanguage Model SteeringMultilingual Language Models