mlsanityData to 7 Oct 2026

Word Embeddings Are Steers for Language Models

arXiv 2305.12798 ↗First posted 22 May 2023cs.CL cs.AI cs.LGcited by 70

Not on the map. Why?

Abstract (arXiv)

Language models (LMs) automatically learn word embeddings during pre-training on language corpora. Although word embeddings are usually interpreted as feature vectors for individual words, their roles in language model generation remain underexplored. In this work, we theoretically and empirically revisit output word embeddings and find that their linear transformations are equivalent to steering language model generation styles. We name such steers LM-Steers and find them existing in LMs of all sizes. It requires learning parameters equal to 0.2% of the original LMs' size for steering each style. On tasks such as language model detoxification and sentiment control, LM-Steers can achieve comparable or superior performance compared with state-of-the-art controlled generation methods while maintaining a better balance with generation quality. The learned LM-Steer serves as a lens in text styles: it reveals that word embeddings are interpretable when associated with language model generations and can highlight text spans that most indicate the style differences. An LM-Steer is transferrable between different language models by an explicit form calculation. One can also continuously steer LMs simply by scaling the LM-Steer or compose multiple LM-Steers by adding their transformations. Our codes are publicly available at urlhttps://github.com/Glaciohound/LM-Steer.

Key facts. Cited by 70 papers in the map (data through 7 Oct 2026). Builds on 37 of its 70 references that are in the map. First posted on arXiv 22 May 2023.

Builds on

37 of 70 references.

References · 37

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2020Language Models are Few-Shot Learnerscited by 32,199
  3. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
  4. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  5. 2013Efficient Estimation of Word Representations in Vector Spacecited by 4,346
  6. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  7. 2021Prefix-Tuning: Optimizing Continuous Prompts for Generationcited by 3,878
  8. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  9. 2023Pythia: A Suite for Analyzing Large Language Models Across Training and Scalingcited by 1,734
  10. 2020Don't Stop Pretraining: Adapt Language Models to Domains and Taskscited by 1,548
  11. 2016Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddingscited by 1,304
  12. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,235
25 more
  1. 2020StereoSet: Measuring stereotypical bias in pretrained language modelscited by 911
  2. 2019Plug and Play Language Models: A Simple Approach to Controlled Text Generationcited by 793
  3. 2020CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Modelscited by 703
  4. 2019The Woman Worked as a Babysitter: On Biases in Language Generationcited by 540
  5. 2020Null It Out: Guarding Protected Attributes by Iterative Nullspace Projectioncited by 442
  6. 2021DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Expertscited by 434
  7. 2020GeDi: Generative Discriminator Guided Sequence Generationcited by 377
  8. 2021FUDGE: Controlled Text Generation With Future Discriminatorscited by 374
  9. 2021Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLPcited by 336
  10. 2019On the "steerability" of generative adversarial networkscited by 247
  11. 2020Measuring and Reducing Gendered Correlations in Pre-trained Modelscited by 222
  12. 2018Delete, Retrieve, Generate: A Simple Approach to Sentiment and Style Transfercited by 191
  13. 2020Towards Debiasing Sentence Representationscited by 170
  14. 2019Queens are Powerful too: Mitigating Gender Bias in Dialogue Generationcited by 152
  15. 2021RedditBias: A Real-World Resource for Bias Evaluation and Debiasing of Conversational Language Modelscited by 149
  16. 2021Debiasing Pre-trained Contextualised Embeddingscited by 107
  17. 2023Controlled Text Generation with Natural Language Instructionscited by 87
  18. 2021Controlled Text Generation as Continuous Optimization with Multiple Constraintscited by 70
  19. 2022Gradient-Based Constrained Sampling from Language Modelscited by 64
  20. 2019Analogies Explained: Towards Understanding Word Embeddingscited by 63
  21. 2022Debiasing isn't enough! -- On the Effectiveness of Debiasing MLMs and their Social Biases in Downstream Taskscited by 48
  22. 2022NormSAGE: Multi-Lingual Multi-Cultural Norm Discovery from Conversations On-the-Flycited by 48
  23. 2017Semantic Structure and Interpretability of Word Embeddingscited by 21
  24. 2019Rotate King to get Queen: Word Relationships as Orthogonal Transformations in Embedding Spacecited by 4
  25. 2020MUSE: Textual Attributes Guided Portrait Painting Generationcited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 3; 2024: 13; 2025: 35; 2026: 19.

70 papers.

Sort:

By area:

  1. 2024Programming Refusal with Conditional Activation Steeringcited by 191
  2. 2025MemOS: A Memory OS for AI Systemcited by 129
  3. 2024Aligning LLMs with Individual Preferences via Interactioncited by 63
  4. 2024Steering Away from Harm: An Adaptive Approach to Defending Vision Language Model Against Jailbreakscited by 44
  5. 2023HallE-Control: Controlling Object Hallucination in Large Multimodal Modelscited by 31
  6. 2024Reefknot: A Comprehensive Benchmark for Relation Hallucination Evaluation, Analysis and Mitigation in Multimodal Large Language Modelscited by 30
  7. 2025SafeSwitch: Steering Unsafe LLM Behavior via Internal Activation Signalscited by 29
  8. 2025SAE-SSV: Supervised Steering in Sparse Representation Spaces for Reliable Control of Language Modelscited by 28
  9. 2024Evaluating the Prompt Steerability of Large Language Modelscited by 26
  10. 2024AutoPRM: Automating Procedural Supervision for Multi-Step Reasoning via Controllable Question Decompositioncited by 24
  11. 2025Beyond Prompt Engineering: Robust Behavior Control in LLMs via Steering Target Atomscited by 23
  12. 2025Steering MoE LLMs via Expert (De)Activationcited by 23
38 more
  1. 2025Automating Steering for Safe Multimodal Large Language Modelscited by 17
  2. 2025Personalize Your LLM: Fake it then Align itcited by 15
  3. 2025SafeScientist: Toward Risk-Aware Scientific Discoveries by LLM Agentscited by 14
  4. 2025EasySteer: A Unified Framework for High-Performance and Extensible LLM Steeringcited by 13
  5. 2025CLASH: Evaluating Language Models on Judging High-Stakes Dilemmas from Multiple Perspectivescited by 10
  6. 2024PropaInsight: Toward Deeper Understanding of Propaganda in Terms of Techniques, Appeals, and Intentcited by 9
  7. 2025Focus Directions Make Your Language Models Pay More Attention to Relevant Contextscited by 9
  8. 2025UniDetox: Universal Detoxification of Large Language Models via Dataset Distillationcited by 8
  9. 2025Breaking Bad Tokens: Detoxification of LLMs Using Sparse Autoencoderscited by 8
  10. 2026Why Steering Works: Toward a Unified View of Language Model Parameter Dynamicscited by 8
  11. 2026ICON: Indirect Prompt Injection Defense for Agents based on Inference-Time Correctioncited by 8
  12. 2024Fake Artificial Intelligence Generated Contents (FAIGC): A Survey of Theories, Detection Methods, and Opportunitiescited by 7
  13. 2025EasyEdit2: An Easy-to-use Steering Framework for Editing Large Language Modelscited by 7
  14. 2024From Word Vectors to Multimodal Embeddings: Techniques, Applications, and Future Directions For Large Language Modelscited by 6
  15. 2025Can LLMs Understand Unvoiced Speech? Exploring EMG-to-Text Conversion with LLMscited by 6
  16. 2026Fine-Grained Activation Steering: Steering Less, Achieving Morecited by 6
  17. 2024Focus On This, Not That! Steering LLMs with Adaptive Feature Specificationcited by 5
  18. 2025Exploring Model Editing for LLM-based Aspect-Based Sentiment Classificationcited by 5
  19. 2025Overcoming Multi-step Complexity in Multimodal Theory-of-Mind Reasoning: A Scalable Bayesian Plannercited by 5
  20. 2025Cache Mechanism for Agent RAG Systemscited by 5
  21. 2025Fantastic Reasoning Behaviors and Where to Find Them: Unsupervised Discovery of the Reasoning Processcited by 5
  22. 2026Compositional Steering of Large Language Models with Steering Tokenscited by 5
  23. 2026EmoShift: Lightweight Activation Steering for Enhanced Emotion-Aware Speech Synthesiscited by 5
  24. 2023Defining a New NLP Playgroundcited by 4
  25. 2025Learning Distribution-Wise Control in Representation Space for Language Modelscited by 4
  26. 2025Continuously Steering LLMs Sensitivity to Contextual Knowledge with Proxy Modelscited by 4
  27. 2026Sparse Autoencoders are Capable LLM Jailbreak Mitigatorscited by 4
  28. 2025Dialz: A Python Toolkit for Steering Vectorscited by 3
  29. 2026CausalDetox: Causal Head Selection and Intervention for Language Model Detoxificationcited by 3
  30. 2024Continuous Language Model Interpolation for Dynamic and Controllable Text Generationcited by 2
  31. 2024LDC: Learning to Generate Research Idea with Dynamic Controlcited by 2
  32. 2025From the New World of Word Embeddings: A Comparative Study of Small-World Lexico-Semantic Networks in LLMscited by 2
  33. 2025LLM Braces: Straightening Out LLM Predictions with Relevant Sub-Updatescited by 2
  34. 2025From Word to World: Evaluate and Mitigate Culture Bias in LLMs via Word Association Testcited by 2
  35. 2025Zero-Shot Vision Encoder Grafting via LLM Surrogatescited by 2
  36. 2025Balancing Stylization and Truth via Disentangled Representation Steeringcited by 2
  37. 2025ThematicPlane: Bridging Tacit User Intent and Latent Spaces for Image Generationcited by 2
  38. 2025On Optimal Steering to Achieve Exact Fairnesscited by 2

50 most cited of 70 All 70 (JSON)