Language Model Steering

Ask about this topic

Overview

Language model steering studies how to make a pretrained model change a particular behavior—such as tone, refusal, truthfulness, reasoning, or language—without necessarily retraining its weights. Common approaches add contrastive activation vectors, target sparse features or specific neurons and attention heads, or adapt interventions to the prompt and the model’s evolving state; other methods guide logits, decoding, or agent-level decisions. Researchers evaluate these methods on tasks ranging from safety and factuality to personalization, multilingual generation, and reasoning, measuring both target behavior and preservation of general capabilities. A central open challenge is reliable, selective control: a behavior may be readable in activations yet resist intervention, while steering can cause collateral quality loss, prompt sensitivity, or safety trade-offs.

Latest papers

CardsList

283 papers51 papers in the last four weeks, up 65% on the four weeks before. 0.3% of all new papers.

Mar 18, 2026

Interpretability without actionability: mechanistic methods cannot correct language model errors despite near-perfect internal representations

Language models encode task-relevant knowledge in internal representations that far exceeds their output performance, but whether mechanistic interpretability methods can bridge this knowledge-action gap has not been systematically tested. We compared four mechanistic interpretability methods -- concept bottleneck steering (Steerling-8B), sparse autoencoder feature steering, logit lens with activation patching, and linear probing with truthfulness separator vector steering (Qwen 2.5 7B Instruct) -- for correcting false-negative triage errors using 400 physician-adjudicated clinical vignettes (144 hazards, 256 benign). Linear probes discriminated hazardous from benign cases with 98.2% AUROC, yet the model's output sensitivity was only 45.1%, a 53-percentage-point knowledge-action gap. Concept bottleneck steering corrected 20% of missed hazards but disrupted 53% of correct detections, indistinguishable from random perturbation (p=0.84). SAE feature steering produced zero effect despite 3,695 significant features. TSV steering at high strength corrected 24% of missed hazards while disrupting 6% of correct detections, but left 76% of errors uncorrected. Current mechanistic interpretability methods cannot reliably translate internal knowledge into corrected outputs, with implications for AI safety frameworks that assume interpretability enables effective error correction.

Sanjay Basu, Sadiq Y. Patel, Parth Sheth +5

University of California San Francisco, San Francisco, CA, USA · Waymark, San Francisco, CA, USA · University of Pennsylvania, Philadelphia, PA, USA +2

Mechanistic InterpretabilityLanguage Model Steering
AskPDF
Mar 18, 2026

FaithSteer-BENCH: A Deployment-Aligned Stress-Testing Benchmark for Inference-Time Steering

Inference-time steering is widely regarded as a lightweight and parameter-free mechanism for controlling large language model (LLM) behavior, and prior work has often suggested that simple activation-level interventions can reliably induce targeted behavioral changes. However, such conclusions are typically drawn under relatively relaxed evaluation settings that overlook deployment constraints, capability trade-offs, and real-world robustness. We therefore introduce FaithSteer-BENCH, a stress-testing benchmark that evaluates steering methods at a fixed deployment-style operating point through three gate-wise criteria: controllability, utility preservation, and robustness. Across multiple models and representative steering approaches, we uncover several systematic failure modes that are largely obscured under standard evaluation, including illusory controllability, measurable cognitive tax on unrelated capabilities, and substantial brittleness under mild instruction-level perturbations, role prompts, encoding transformations, and data scarcity. Gate-wise benchmark results show that existing methods do not necessarily provide reliable controllability in deployment-oriented practical settings. In addition, mechanism-level diagnostics indicate that many steering methods induce prompt-conditional alignment rather than stable latent directional shifts, further explaining their fragility under stress. FaithSteer-BENCH therefore provides a unified benchmark and a clearer analytical lens for future method design, reliability evaluation, and deployment-oriented research in steering.

Zikang Ding, Qiying Hu, Yi Zhang +4

University of Electronic Science and Technology of China · Mohamed bin Zayed University of Artificial Intelligence · South China University of Technology

Language Model SteeringLanguage Model Robustness
AskPDF
Mar 18, 2026

A Unified Language Model for Large Scale Search, Recommendation, and Reasoning

LLMs are increasingly applied to recommendation, retrieval, and reasoning, yet deploying a single end-to-end model that can jointly support these behaviors over large, heterogeneous catalogs remains challenging. Such systems must generate unambiguous references to real items, handle multiple entity types, and operate under strict latency and reliability constraints requirements that are difficult to satisfy with text-only generation. While tool-augmented recommender systems address parts of this problem, they introduce orchestration complexity and limit end-to-end optimization. We view this setting as an instance of a broader research problem: how to adapt LLMs to reason jointly over multiple-domain entities, users, and language in a fully self-contained manner. To this end, we introduce NEO, a framework that adapts a pre-trained decoder-only LLM into a tool-free, catalog-grounded generator. NEO represents items as SIDs and trains a single model to interleave natural language and typed item identifiers within a shared sequence. Text prompts control the task, target entity type, and output format (IDs, text, or mixed), while constrained decoding guarantees catalog-valid item generation without restricting free-form text. We refer to this instruction-conditioned controllability as language-steerability. We treat SIDs as a distinct modality and study design choices for integrating discrete entity representations into LLMs via staged alignment and instruction tuning. We evaluate NEO at scale on a real-world catalog of over 10M items across multiple media types and discovery tasks, including recommendation, search, and user understanding. In offline experiments, NEO consistently outperforms strong task-specific baselines and exhibits cross-task transfer, demonstrating a practical path toward consolidating large-scale discovery capabilities into a single language-steerable generative model.

Marco De Nadai, Edoardo D'Amico, Max Lefarov +18

Spotify Denmark, Spain, Germany, France, United States, United Kingdom

Large Language Model-Based RecommendationLLM GroundingLanguage Model Steering
AskPDF
Mar 18, 2026

SaFeR-Steer: Evolving Multi-Turn MLLMs via Synthetic Bootstrapping and Feedback Dynamics

MLLMs are increasingly deployed in multi-turn settings, where attackers can escalate unsafe intent through the evolving visual-text history and exploit long-context safety decay. Yet safety alignment is still dominated by single-turn data and fixed-template dialogues, leaving a mismatch between training and deployment. To bridge this gap, we propose SaFeR-Steer, a progressive multi-turn alignment framework that combines staged synthetic bootstrapping with tutor-in-the-loop GRPO to train a single student under adaptive, on-policy attacks. We also introduce Trajectory-Consistent Summative Reward (TCSR), which aggregates the historical minimum and average of turn rewards so that any low-quality turn affects the trajectory-level return. I. Dataset. We release STEER, a multi-turn multimodal safety dataset with STEER-SFT (12,934), STEER-RL (2,000), and STEER-Bench (3,227) dialogues spanning 1-10 turns. II. Experiment. Starting from Qwen2.5-VL-3B/7B, SaFeR-Steer substantially improves Safety/Helpfulness on both single-turn (48.30/45.86 →\rightarrow 81.84/70.77 for 3B; 56.21/60.32 →\rightarrow 87.89/77.40 for 7B) and multi-turn benchmarks (12.55/27.13 →\rightarrow 55.58/70.27 for 3B; 24.66/46.48 →\rightarrow 64.89/72.35 for 7B), shifting failures to later turns and yielding robustness beyond scaling alone. Code is available at https://github.com/Ed-Bg/SaFeR-Steer-full

Haolong Hu, Hanyu Li, Tiancheng He +6

Huazhong University of Science and Technology · Beijing University of Posts and Telecommunications · West China Biomedical Big Data Center, Sichuan University +2

Language Model SteeringMultimodal Large Language ModelsLLM Safety AlignmentRL for VLMsMultimodal Multi-Turn Safety Alignment
AskPDF
Mar 17, 2026

Behavioral Steering in a 35B MoE Language Model via SAE-Decoded Probe Vectors: One Agency Axis, Not Five Traits

We train nine sparse autoencoders (SAEs) on the residual stream of Qwen 3.5-35B-A3B, a 35-billion-parameter Mixture-of-Experts model with a hybrid GatedDeltaNet/attention architecture, and use them to identify and steer five agentic behavioral traits. Our method trains linear probes on SAE latent activations, then projects the probe weights back through the SAE decoder to obtain continuous steering vectors in the model's native activation space. This bypasses the SAE's top-k discretization, enabling fine-grained behavioral intervention at inference time with no retraining. Across 1,800 agent rollouts (50 scenarios times 36 conditions), we find that autonomy steering at multiplier 2 achieves Cohen's d = 1.01 (p < 0.0001), shifting the model from asking the user for help 78% of the time to proactively executing code and searching the web. Cross-trait analysis, however, reveals that all five steering vectors primarily modulate a single dominant agency axis (the disposition to act independently versus defer to the user), with trait specific effects appearing only as secondary modulations in tool-type composition and dose-response shape. The tool-use vector steers behavior (d = 0.39); the risk-calibration vector produces only suppression. We additionally show that steering only during autoregressive decoding has zero effect (p > 0.35), providing causal evidence that behavioral commitments are computed during prefill in GatedDeltaNet architectures.

Jia Qing Yap

Independent Researcher

Language Model SteeringCausal Interventions in Language ModelsLLM Agents
AskPDF
Mar 17, 2026

SpecSteer: Synergizing Local Context and Global Reasoning for Efficient Personalized Generation

Realizing personalized intelligence faces a core dilemma: sending user history to centralized large language models raises privacy concerns, while on-device small language models lack the reasoning capacity required for high-quality generation. Our pilot study shows that purely local enhancements remain insufficient to reliably bridge this gap. We therefore propose SpecSteer, an asymmetric collaborative inference framework that synergizes private on-device context with cloud-scale reasoning. SpecSteer casts collaboration as Bayesian knowledge fusion and repurposes speculative decoding as a distributed alignment protocol, yielding a Draft--Verify--Recover pipeline: the on-device model drafts personalized sequences; the cloud validates via a ratio-based mechanism that decouples reasoning verification from private context, filtering logical flaws without accessing raw user context; upon rejection, a steering recovery injects local intent during correction. Experiments demonstrate that SpecSteer successfully closes the reasoning gap and achieves superior personalized generation performance, while delivering a 2.36x speedup over standard baselines.

Hang Lv, Sheng Liang, Hao Wang +6

University of Science and Technology of China · Huawei Technologies Co., Ltd.

Personalized Language ModelsOn-Device Language Model InferencePrivacy-Preserving Language ModelsEfficient Language Model InferenceLanguage Model Steering
AskPDF
Mar 15, 2026

Gradient Atoms: Unsupervised Discovery, Attribution and Steering of Model Behaviors via Sparse Decomposition of Training Gradients

Training data attribution (TDA) methods ask which training documents are responsible for a model behavior. However, models often learn broad concepts shared across many examples. Moreover, existing TDA methods are supervised -- they require a predefined query behavior, then score every training document against it -- making them both expensive and unable to surface behaviors the user did not think to ask about. We present Gradient Atoms, an unsupervised method that decomposes per-document training gradients into sparse components ("atoms") via dictionary learning in a preconditioned eigenspace. Each atom captures a shared update direction induced by a cluster of functionally similar documents, directly recovering the collective structure that per-document methods do not address. Among 500 discovered atoms, the highest-coherence ones recover interpretable task-type behaviors -- refusal, arithmetic, yes/no classification, trivia QA -- without any behavioral labels. These atoms double as effective steering vectors: applying them as weight-space perturbations produces large, controllable shifts in model behavior (e.g., bulleted-list generation 33% to 94%; systematic refusal 50% to 0%). The method requires no query--document scoring stage, and scales independently of the number of query behaviors of interest. Code is available at https://github.com/jrosseruk/gradient_atoms.

J Rosser

FLAIR, University of Oxford

Gradient-Based AttributionTraining Data AttributionLanguage Model Steering
AskPDF
Mar 15, 2026

Nudging Hidden States: Training-Free Model Steering for Chain-of-Thought Reasoning in Large Audio-Language Models

Chain-of-thought (CoT) prompting has been extended to large audio-language models (LALMs) to elicit reasoning, yet enhancing its effectiveness without training remains challenging. We study inference-time model steering as a training-free approach to improve LALM reasoning. We introduce three strategies using diverse information sources and evaluate them across four LALMs and four benchmarks. Results show general accuracy gains up to 4.4% over CoT prompting. Notably, we identify a cross-modal transfer where steering vectors derived from few text samples effectively guide speech-based reasoning, demonstrating high data efficiency. We also examine hyperparameter sensitivity to understand the robustness of these approaches. Our findings position model steering as a practical direction for strengthening LALM reasoning.

Lok-Lam Ieong, Chia-Chien Chen, Chih-Kai Yang +3

National Taiwan University, Taiwan

Language Model SteeringAudio-Language Models
AskPDF
Mar 12, 2026

Global Evolutionary Steering: Refining Activation Steering Control via Cross-Layer Consistency

Activation engineering enables precise control over Large Language Models (LLMs) without the computational cost of fine-tuning. However, existing methods deriving vectors from static activation differences are susceptible to high-dimensional noise and layer-wise semantic drift, often capturing spurious correlations rather than the target intent. To address this, we propose Global Evolutionary Refined Steering (GER-steer), a training-free framework that grounded in the geometric stability of the network's representation evolution. GER-steer exploits this global signal to rectify raw steering vectors, effectively decoupling robust semantic intent from orthogonal artifacts. Extensive evaluations confirm that GER-steer consistently outperforms baselines, delivering superior efficacy and generalization without layer-specific tuning, establishing a universal solution for reliable model alignment.

Xinyan Jiang, Wenjing Yu, Di Wang +1

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) · Provable Responsible AI and Data Analytics (PRADA) Lab · King Abdullah University of Science and Technology +3

Language Model SteeringActivation SteeringLLM Alignment
AskPDF
Mar 11, 2026

PRISM-ΔΔ: Differential Subspace Steering for Prompt Highlighting in Large Language Models

Prompt highlighting steers a large language model to prioritize user-specified text spans during generation. A key challenge of existing Key-editing approaches is extracting steering directions that capture the difference between relevant and irrelevant contexts, rather than shared structural patterns common to both. We propose PRISM-ΔΔ (Projection-based Relevance-Informed Steering Method), which decomposes the difference between positive and negative cross-covariance matrices to maximize discriminative energy while eliminating shared directions. Each attention head receives a continuous softplus importance weight, letting weak-but-useful heads contribute at reduced strength. The framework extends naturally to Value representations, capturing content-channel signal that Key-only methods leave unused. Across four benchmarks and five models, our methods match or exceed the best existing method on 19 of 20 configurations, with relative gains up to +13.0%, while halving the fluency cost of steering. Our methods also scale to long-context retrieval, outperforming the best existing method by up to +5.1% relative gain. PRISM-ΔΔ is compatible with FlashAttention and adds negligible memory overhead.

Yuyao Ge, Shenghua Liu, Yiwei Wang +5

♣Institute of Computing Technology, Chinese Academy of Sciences · △University of California, Merced

Language Model SteeringLLM PromptingActivation Steering
AskPDF
Mar 10, 2026

Curveball Steering: The Right Direction To Steer Isn't Always Linear

Activation steering is a widely used approach for controlling large language model (LLM) behavior by intervening on internal representations. Existing methods largely rely on the Linear Representation Hypothesis, assuming behavioral attributes can be manipulated using global linear directions. In practice, however, such linear interventions often behave inconsistently. We question this assumption by analyzing the intrinsic geometry of LLM activation spaces. Measuring geometric distortion via the ratio of geodesic to Euclidean distances, we observe substantial and concept-dependent distortions, indicating that activation spaces are not well-approximated by a globally linear geometry. Motivated by this, we propose "Curveball steering", a nonlinear steering method based on polynomial kernel PCA that performs interventions in a feature space, better respecting the learned activation geometry. Curveball steering consistently outperforms linear PCA-based steering, particularly in regimes exhibiting strong geometric distortion, suggesting that geometry-aware, nonlinear steering provides a principled alternative to global, linear interventions.

Shivam Raval, Hae Jin Song, Linlin Wu +4

Language Model SteeringActivation SteeringRepresentation Geometry in Language ModelsPrincipal Component Analysis
AskPDF
Mar 9, 2026

How Emotion Shapes the Behavior of LLMs and Agents: A Mechanistic Study

Emotion plays an important role in human cognition and performance. Motivated by this, we investigate whether analogous emotional signals can shape the behavior of large language models (LLMs) and agents. Existing emotion-aware studies mainly treat emotion as a surface-level style factor or a perception target, overlooking its mechanistic role in task processing. To address this limitation, we propose E-STEER, an interpretable emotion steering framework that enables direct representation-level intervention in LLMs and agents. It embeds emotion as a structured, controllable variable in hidden states, and with it, we examine the impact of emotion on objective reasoning, subjective generation, safety, and multi-step agent behaviors. The results reveal non-monotonic emotion-behavior relations consistent with established psychological theories, and show that specific emotions not only enhance LLM capability but also improve safety, and systematically shape multi-step agent behaviors.

Moran Sun, Tianlin Li, Yuwei Zheng +4

School of Computer Science and Engineering, Beihang University, Beijing, China · College of Computing and Data Science, Nanyang Technological University, Singapore

Emotion Representation in Language ModelsLanguage Model SteeringAffective ComputingLLM Agent Safety
AskPDF
Mar 9, 2026

From Refusal Tokens to Refusal Control: Discovering and Steering Category-Specific Refusal Directions

Language models are commonly fine-tuned for safety alignment to refuse harmful prompts. One approach fine-tunes them to generate categorical refusal tokens that distinguish different refusal types before responding. In this work, we leverage a version of Llama 3 8B fine-tuned with these categorical refusal tokens to enable inference-time control over fine-grained refusal behavior, improving both safety and reliability. We show that refusal token fine-tuning induces separable, category-aligned directions in the residual stream, which we extract and use to construct categorical steering vectors with a lightweight probe that determines whether to steer toward or away from refusal during inference. In addition, we introduce a learned low-rank combination that mixes these category directions in a whitened, orthonormal steering basis, resulting in a single controllable intervention under activation-space anisotropy, and show that this intervention is transferable across same-architecture model variants without additional training. Across benchmarks, both categorical steering vectors and the low-rank combination consistently reduce over-refusals on benign prompts while increasing refusal rates on harmful prompts, highlighting their utility for multi-category refusal control.

Rishab Alagharu, Ishneet Sukhvinder Singh, Shaibi Shamsudeen +2

Algoverse AI Research · Carnegie Mellon University · University of Maryland

Language Model SteeringLLM Refusal BehaviorCausal Interventions in Language ModelsRefusal Behavior in Language ModelsRefusal Mechanisms in Language Models
AskPDF
Mar 8, 2026

AI Steerability 360: A Toolkit for Steering Large Language Models

The AI Steerability 360 toolkit is an extensible, open-source Python library for steering LLMs. Steering abstractions are designed around four model control surfaces: input (modification of the prompt), structural (modification of the model's weights or architecture), state (modification of the model's activations and attentions), and output (modification of the decoding or generation process). Steering methods exert control on the model through a common interface, termed a steering pipeline, which additionally allows for the composition of multiple steering methods. Comprehensive evaluation and comparison of steering methods/pipelines is facilitated by use case classes (for defining tasks) and a benchmark class (for performance comparison on a given task). The functionality provided by the toolkit significantly lowers the barrier to developing and comprehensively evaluating steering methods. The toolkit is Hugging Face native and is released under an Apache 2.0 license at https://github.com/IBM/AISteer360.

Erik Miehling, Karthikeyan Natesan Ramamurthy, Praveen Venkateswaran +10

IBM Research

Language Model Steering
AskPDF
Mar 6, 2026

COLD-Steer: Steering Large Language Models via In-Context One-step Learning Dynamics

Activation steering methods enable inference-time control of large language model (LLM) behavior without retraining, but current approaches face a fundamental trade-off: sample-efficient methods suboptimally capture steering signals from labeled examples, while methods that better extract these signals require hundreds to thousands of examples. We introduce COLD-Steer, a training-free framework that steers LLM activations by approximating the representational changes that would result from gradient descent on in-context examples. Our key insight is that the effect of fine-tuning on a small set of examples can be efficiently approximated at inference time without actual parameter updates. We formalize this through two complementary approaches: (i) a unit kernel approximation method that updates the activations directly using gradients with respect to them, normalized across examples, and (ii) a finite-difference approximation requiring only two forward passes regardless of example count. Experiments across a variety of steering tasks and benchmarks demonstrate that COLD-Steer achieves upto 95% steering effectiveness while using 50 times fewer samples compared to the best baseline. COLD-Steer facilitates accommodating diverse perspectives without extensive demonstration data, which we validate through our experiments on pluralistic alignment tasks. Our framework opens new possibilities for adaptive, context-aware model control that can flexibly address varying loss-driven human preferences through principled approximation of learning dynamics rather than specialized training procedures.

Kartik Sharma, Rakshit S. Trivedi

Georgia Institute of Technology · Massachusetts Institute of Technology

Language Model SteeringInference-Time Language Model AlignmentActivation Steering
AskPDF
Mar 6, 2026

Enhancing Instruction Following of LLMs via Activation Steering with Dynamic Rejection

Large Language Models (LLMs), despite advances in instruction tuning, often fail to follow complex user instructions. Activation steering techniques aim to mitigate this by manipulating model internals, but have a potential risk of oversteering, where excessive emphasis on the instruction degrades task accuracy and overall text quality. To address this, we introduce DIRECTER (Dynamic rejection steering), a novel steering method that dynamically modulates steering strength by scaling the KV cache without extra dataset. DIRECTER couples steering with a plausibility-guided decoding loop, which adaptively adjusts steering strength at each step by comparing the steered output distribution to the original. If the steered output is deemed implausible, steering strength is progressively weakened. This strength modulation is guided by a lightweight, one-time attention sensitivity analysis that ranks layers by their influence on model representations. Extensive evaluations show that DIRECTER significantly enhances instruction-following capabilities across diverse benchmarks, improving accuracy by up to 6.5% over baselines without the common trade-offs in generation quality or task fidelity. The proposed dynamic, plausibility-guided control during activation steering further demonstrates its potential as a general mechanism for mitigating oversteering that is compatible with existing baselines.

Minjae Kang, Jaehyung Kim

Yonsei University

Language Model SteeringInstruction FollowingActivation SteeringLanguage Model Decoding
AskPDF
Mar 3, 2026

Understanding and Mitigating Dataset Corruption in LLM Steering

Contrastive steering has been shown as a simple and effective method to adjust the generative behavior of LLMs at inference time. It uses examples of prompt responses with and without a trait to identify a direction in an intermediate activation layer, and then shifts activations in this 1-dimensional subspace. However, despite its growing use in AI safety applications, the robustness of contrastive steering to noisy or adversarial data corruption is poorly understood. We initiate a study of the robustness of this process with respect to corruption of the dataset of examples used to train the steering direction. Our first observation is that contrastive steering is quite robust to a moderate amount of corruption, but unwanted side effects can be clearly and maliciously manifested when a non-trivial fraction of the training data is altered. Second, we analyze the geometry of various types of corruption, and identify some safeguards. Notably, a key step in learning the steering direction involves high-dimensional mean computation, and we show that replacing this step with a recently developed robust mean estimator often mitigates most of the unwanted effects of malicious corruption.

Cullen Anderson, Narmeen Oozeer, Foad Namjoo +3

Department of Computer Science, University of Massachusetts Amherst · Martian AI · Kahlert School of Computing, University of Utah +1

Robust Mean EstimationLanguage Model SteeringData Poisoning Attacks
AskPDF
Mar 3, 2026

How Controllable Are Large Language Models? A Unified Evaluation across Behavioral Granularities

Large Language Models (LLMs) are increasingly deployed in socially sensitive domains, yet their unpredictable behaviors, ranging from misaligned intent to inconsistent personality, pose significant risks. We introduce SteerEval, a hierarchical benchmark for evaluating LLM controllability across three domains: language features, sentiment, and personality. Each domain is structured into three specification levels: L1 (what to express), L2 (how to express), and L3 (how to instantiate), connecting high-level behavioral intent to concrete textual output. Using SteerEval, we systematically evaluate contemporary steering methods, revealing that control often degrades at finer-grained levels. Our benchmark offers a principled and interpretable framework for safe and controllable LLM behavior, serving as a foundation for future research.

Ziwen Xu, Kewei Xu, Haoming Xu +8

Zhejiang University · Alibaba Group

LLM EvaluationLanguage Model Steering
AskPDF
Mar 1, 2026

Spectral Attention Steering for Prompt Highlighting

Attention steering is an important technique for controlling model focus, enabling capabilities such as prompt highlighting, where the model prioritises user-specified text. However, existing attention steering methods require explicit storage of the full attention matrix, making them incompatible with memory-efficient implementations like FlashAttention. We introduce Spectral Editing Key Amplification (SEKA), a training-free steering method that tackles this by directly editing key embeddings before attention computation. SEKA uses spectral decomposition to steer key embeddings towards latent directions that amplify attention scores for certain tokens. We extend this to Adaptive SEKA (AdaSEKA), a query-adaptive variant that uses a training-free routing mechanism to dynamically combine multiple expert subspaces based on the prompt's semantic intent. Our experiments show both methods significantly outperform strong baselines on standard steering benchmarks while adding much lower latency and memory overhead, in compatibility with optimised attention.

Weixian Waylon Li, Yuchen Niu, Yongxin Yang +3

University of Edinburgh, UK · RayNeo, China · Queen Mary University of London, UK +1

Language Model SteeringEfficient Attention
AskPDF
Feb 26, 2026

Interpreting and Steering State-Space Models via Activation Subspace Bottlenecks

State-space models (SSMs) have emerged as an efficient strategy for building powerful language models, avoiding the quadratic complexity of computing attention in transformers. Despite their promise, the interpretability and steerability of modern SSMs remain relatively underexplored. We take a major step in this direction by identifying activation subspace bottlenecks in the Mamba family of SSM models using tools from mechanistic interpretability. We then introduce a test-time steering intervention that simply multiplies the activations of the identified bottlenecks by a scalar. Across 7 SSMs and 6 diverse benchmarks, this intervention improves performance by an average of 8.27%, without requiring any task-specific tuning. Finally, we validate that the identified bottlenecks are indeed hindering performance by modifying them to yield an architecture we call Stable-Mamba, which achieves long-context performance gains when retrained from scratch.

Vamshi Sunku Mohan, Kaustubh Gupta, Aneesha Das +1

AppViewX, Bangalore · Independent Researcher · Microsoft Research, Redmond

State Space Language ModelsState Space ModelsLanguage Model SteeringLLM InterpretabilitySelective SSMsActivation Steering
AskPDF
Feb 24, 2026

Steering at the Source: Style Modulation Heads for Robust Persona Control

Activation steering offers a computationally efficient mechanism for controlling Large Language Models (LLMs) without fine-tuning. While effectively controlling target traits (e.g., persona), coherency degradation remains a major obstacle to safety and practical deployment. We hypothesize that this degradation stems from intervening on the residual stream, which indiscriminately affects aggregated features and inadvertently amplifies off-target noise. In this work, we identify a sparse subset of attention heads (only three heads) that independently govern persona and style formation, which we term Style Modulation Heads. Specifically, these heads can be localized via geometric analysis of internal representations, combining layer-wise cosine similarity and head-wise contribution scores. We demonstrate that intervention targeting only these specific heads achieves robust behavioral control while significantly mitigating the coherency degradation observed in residual stream steering. More broadly, our findings show that precise, component-level localization enables safer and more precise model control.

Yoshihiro Izawa, Gouki Minegishi, Koshi Eguchi +2

The University of Tokyo, Tokyo, Japan · National Institute of Informatics Research and Development Center for Large Language Models, Japan

Language Model SteeringCausal Interventions in Language Models
AskPDF
Feb 19, 2026

Understanding Unreliability of Steering Vectors in Language Models: Geometric Predictors and the Limits of Linear Approximations

Steering vectors are a lightweight method for controlling language model behavior by adding a learned bias to the activations at inference time. Although effective on average, steering effect sizes vary across samples and are unreliable for many target behaviors. In my thesis, I investigate why steering reliability differs across behaviors and how it is impacted by steering vector training data. First, I find that higher cosine similarity between training activation differences predicts more reliable steering. Second, I observe that behavior datasets where positive and negative activations are better separated along the steering direction are more reliably steerable. Finally, steering vectors trained on different prompt variations are directionally distinct, yet perform similarly well and exhibit correlated efficacy across datasets. My findings suggest that steering vectors are unreliable when the latent target behavior representation is not effectively approximated by the linear steering direction. Taken together, these insights offer a practical diagnostic for steering unreliability and motivate the development of more robust steering methods that explicitly account for non-linear latent behavior representations.

Joschka Braun

Eberhard Karls University of Tübingen Faculty of Science

Language Model SteeringActivation SteeringLLM ReliabilityRepresentation Geometry in Language Models
AskPDF
Feb 19, 2026

ODESteer: A Unified ODE-Based Steering Framework for LLM Alignment

Activation steering, or representation engineering, offers a lightweight approach to align large language models (LLMs) by manipulating their internal activations at inference time. However, current methods suffer from two key limitations: (i) the lack of a unified theoretical framework for guiding the design of steering directions, and (ii) an over-reliance on one-step steering that fail to capture complex patterns of activation distributions. In this work, we propose a unified ordinary differential equations (ODEs)-based theoretical framework for activation steering in LLM alignment. We show that conventional activation addition can be interpreted as a first-order approximation to the solution of an ODE. Based on this ODE perspective, identifying a steering direction becomes equivalent to designing a barrier function from control theory. Derived from this framework, we introduce ODESteer, a kind of ODE-based steering guided by barrier functions, which shows empirical advancement in LLM alignment. ODESteer identifies steering directions by defining the barrier function as the log-density ratio between positive and negative activations, and employs it to construct an ODE for multi-step and adaptive steering. Compared to state-of-the-art activation steering methods, ODESteer achieves consistent empirical improvements on diverse LLM alignment benchmarks, a notable 5.7%5.7\% improvement over TruthfulQA, 2.5%2.5\% over UltraFeedback, and 2.4%2.4\% over RealToxicityPrompts. Our work establishes a principled new view of activation steering in LLM alignment by unifying its theoretical foundations via ODEs, and validating it empirically through the proposed ODESteer method.

Hongjue Zhao, Haosen Sun, Jiangtao Kong +8

University of Illinois Urbana-Champaign · Northwestern University · William & Mary +2

Language Model SteeringActivation SteeringLLM AlignmentInference-Time Language Model Alignment
AskPDF
Feb 17, 2026

Activation Steering via Generative Causal Mediation

Where should we intervene in a language model (LM) to localize and control behaviors that are diffused across many tokens of a long-form response? We introduce Generative Causal Mediation (GCM), a procedure for selecting model components (e.g., attention heads) from contrastive long-form responses, to steer such diffuse concepts (e.g., talk in verse vs. talk in prose). In GCM, we first construct a dataset of contrasting behavioral inputs and long-form responses. Then, we quantify how model components mediate the concept and select the strongest mediators for steering. We evaluate GCM on three behaviors--refusal, sycophancy, and style transfer--across three language models. GCM successfully localizes concepts expressed in long-form responses and outperforms correlational probe-based baselines when steering with a sparse set of attention heads. Together, these results demonstrate that GCM provides an effective approach for localizing from and controlling the long-form responses of LMs.

Aruna Sankaranarayanan, Amir Zur, Atticus Geiger +1

CSAIL, MIT · Pr(AI)2R · Stanford University +1

Language Model SteeringCausal Interventions in Language ModelsCausal Interpretability
AskPDF
Feb 17, 2026

The Information Geometry of Softmax: Probing and Steering

This paper concerns the question of how AI systems encode semantic structure into the geometric structure of their representation spaces. The motivating observation is that the natural geometry of these representation spaces should reflect the way models use representations to produce behavior. We focus on the important special case of representations that define softmax distributions. In this case, we argue that the natural geometry is information geometry. Our focus is on the role of information geometry on semantic encoding and the linear representation hypothesis. As an illustrative application, we develop "dual steering", a method for robustly steering representations to exhibit a particular concept using linear probes. We prove that dual steering optimally modifies the target concept while minimizing changes to off-target concepts. Empirically, we find that dual steering enhances the controllability and stability of concept manipulation.

Kiho Park, Todd Nief, Yo Joong Choe +1

University of Chicago · INSEAD

Representation Geometry in Language ModelsInformation GeometryRepresentation SteeringLanguage Model SteeringLinear Probing
AskPDF
Feb 15, 2026

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

When a model solves a problem on one attempt and fails it on the next, what separates the two is rarely the final answer token; it is the trajectory that reached it. Contrastive activation steering leaves that signal unused: CAA, SADI, RepE and ITI build their direction from experimenter-supplied text, recorded while the model reads rather than reasons. That choice also caps what the vector can express, since polarity must be written into the text, and a task judged only by outcome offers nothing to write it with. ROAST makes the trajectory itself the contrast: sample rollouts, let an outcome verifier split them into successes and failures, and contrast the reasoning that worked against the reasoning that did not. A matched teacher-forced control---rollouts, labels, answer text and pair counts held fixed, the trajectory alone stripped---points to the trajectory as what matters: on GSM8K at 0.6B the pairs alone buy +0.12 points while restoring the trajectories buys +6.05, the larger and only seed-robust step. Replacing the trajectory with an equal-length neutral prefix or another question's reasoning falls below no intervention. The two corpora are also far apart geometrically, a median 70+ degrees apart at both Qwen3 scales probed, beyond what a split-half null explains. Reading from rollouts calls for two corrections---keeping the full difference vector rather than Top-10% masking, and giving each question one vote rather than one per pair---and only grouped aggregation beats the unsteered baseline under 20% verifier noise. On parser-free benchmarks (GSM8K, MATH500, IFEval), ROAST is best in all six cells over two models, by up to +9.7, at +6.4% wall-clock and no added context; it also leads on six parser-scored benchmarks across three models. Across nine models (0.6B--122B, four families), ROAST improves on the unsteered model at every scale. Code: https://github.com/TomySu404/ORBIT

Xuanbo Su, Yingfang Zhang, Hao Luo +3

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

Language Model SteeringRepresentation SteeringActivation SteeringReasoning Trace Analysis
AskPDF
Feb 6, 2026

Endogenous Resistance to Activation Steering in Language Models

Large language models can recover mid-generation from task-misaligned activation steering, producing explicit verbal restarts (e.g., ``wait, that's not right'') and continuing on-topic even while the steering perturbation remains active. We term this Endogenous Steering Resistance (ESR). Using sparse autoencoder (SAE) latents to steer model activations, we find that Llama-3.3-70B exhibits explicit ESR at 3.8%, with smaller models from the Llama-3 and Gemma-2 families showing the explicit form less frequently. Two controls dissociate ESR into a detection event and a sustained-resistance component that conditioning on recent on-topic tokens does not fully explain. We identify 26 SAE latents through contrastive on-topic/off-topic search; zero-ablating them reduces the multi-attempt rate by 25%, with random-latent and held-out-prompt controls supporting specificity. ESR can also be deliberately enhanced through both meta-prompting and fine-tuning on synthetic self-correction examples. ESR has dual implications for safety: it could harden models against adversarial activation-space manipulation, but may equally interfere with beneficial steering-based interventions, since the model has no way to distinguish the two. Code is available at https://github.com/agencyenterprise/endogenous-steering-resistance.

Alex McKenzie, Keenan Pepper, Stijn Servaes +6

AE Studio · Princeton Neuroscience Institute & Department of Psychology, Princeton University, Princeton, NJ.

Language Model SteeringLanguage Model RobustnessActivation Steering
AskPDF
Feb 2, 2026

Towards Understanding Steering Strength

A popular approach to post-training control of large language models (LLMs) is the steering of intermediate latent representations. Namely, identify a well-chosen direction depending on the task at hand and perturbs representations along this direction at inference time. While many propositions exist to pick this direction, considerably less is understood about how to choose the magnitude of the move, whereas its importance is clear: too little and the intended behavior does not emerge, too much and the model's performance degrades beyond repair. In this work, we propose the first theoretical analysis of steering strength. We characterize its effect on next token probability, presence of a concept, and cross-entropy, deriving precise qualitative laws governing these quantities. Our analysis reveals surprising behaviors, including non-monotonic effects of steering strength. We validate our theoretical predictions empirically on eleven language models, ranging from a small GPT architecture to modern models.

Magamed Taimeskhanov, Samuel Vaiter, Damien Garreau

University of Würzburg, Center for Artificial Intelligence and Data Science · Universit´e Cˆote d’Azur - CNRS

Language Model SteeringActivation Steering
AskPDF
Feb 2, 2026

From Directions to Regions: Decomposing Activations in Language Models via Local Geometry

Activation decomposition methods in language models are tightly coupled to geometric assumptions on how concepts are realized in activation space. Existing approaches search for individual global directions, implicitly assuming linear separability, which overlooks concepts with nonlinear or multi-dimensional structure. In this work, we leverage Mixture of Factor Analyzers (MFA) as a scalable, unsupervised alternative that models the activation space as a collection of Gaussian regions with their local covariance structure. MFA decomposes activations into two compositional geometric objects: the region's centroid in activation space, and the local variation from the centroid. We train large-scale MFAs for Llama-3.1-8B and Gemma-2-2B, and show they capture complex, nonlinear structures in activation space. Moreover, evaluations on localization and steering benchmarks show that MFA outperforms unsupervised baselines, is competitive with supervised localization methods, and often achieves stronger steering performance than sparse autoencoders. Together, our findings position local geometry, expressed through subspaces, as a promising unit of analysis for scalable concept discovery and model control, accounting for complex structures that isolated directions fail to capture.

Or Shafran, Shaked Ronen, Omri Fahn +3

Blavatnik School of Computer Science and AI, Tel Aviv University, Israel · New York University, New York, NY, USA · Goodfire.

Language Model SteeringLLM InterpretabilityGaussian Mixture ModelsLanguage ModelingRepresentation Geometry in Language ModelsConcept Localization in Language Models
AskPDF
Jan 29, 2026

ILRR: Inference-Time Steering Method for Masked Diffusion Language Models

Discrete Diffusion Language Models (DLMs) offer a promising non-autoregressive alternative for text generation, yet effective mechanisms for inference-time control remain relatively underexplored. Existing approaches include sampling-level guidance or trajectory optimization mechanisms. In this work, we study the paradigm of reference-based latent steering for DLMs. We introduce Iterative Latent Representation Refinement (ILRR), an efficient framework for steering DLMs using a reference text as a high-level semantic blueprint. ILRR extracts and injects reference-derived semantic signals into the evolving activations of the generated sequence, enabling tunable transfer of coarse properties such as sentiment. We further introduce Spatially Modulated Steering, an extension that enables long-form generation to be guided by shorter references by regulating intensity across the sequence. Empirically, we demonstrate that ILRR achieves effective control on LLaDA and MDLM architectures with low computational overhead, requiring only one additional parallel forward pass per denoising step. Under comparable compute budgets, ILRR improves attribute accuracy over baselines by 10% to 60% points. Our results suggest that the iterative, global denoising process makes DLMs a natural substrate for effective sequence-wide activation-level control.

Eden Avrahami, Eliya Nachmani

School of Computer Science, Tel Aviv University, Israel · Department of Electrical and Computer Engineering, Ben Gurion University, Beer Sheva, Israel.

Masked Diffusion ModelsLanguage Model SteeringControllable Text GenerationDiffusion Language Model InferenceDiffusion Language ModelsActivation Steering
AskPDF