mlsanityData to 14 Oct 2026

Controlling Language and Diffusion Models by Transporting Activations

arXiv 2410.23054 ↗First posted 30 Oct 2024cs.LG cs.AI cs.CLcited by 68

On the map: Diffusion models → Discrete and text diffusion

Abstract (arXiv)

The increasing capabilities of large generative models and their ever more widespread deployment have raised concerns about their reliability, safety, and potential misuse. To address these issues, recent works have proposed to control model generation by steering model activations in order to effectively induce or prevent the emergence of concepts or behaviors in the generated output. In this paper we introduce Activation Transport (AcT), a general framework to steer activations guided by optimal transport theory that generalizes many previous activation-steering works. AcT is modality-agnostic and provides fine-grained control over the model behavior with negligible computational overhead, while minimally impacting model abilities. We experimentally show the effectiveness and versatility of our approach by addressing key challenges in large language models (LLMs) and text-to-image diffusion models (T2Is). For LLMs, we show that AcT can effectively mitigate toxicity, induce arbitrary concepts, and increase their truthfulness. In T2Is, we show how AcT enables fine-grained style control and concept negation.

Key facts. Cited by 68 papers in the map (data through 14 Oct 2026). Placed in Discrete and text diffusion, one of 14 directions in Diffusion models. Builds on 51 of its 61 references that are in the map. First posted on arXiv 30 Oct 2024.

Builds on

51 of 61 references.

References · 51

  1. 2017Attention Is All You Needcited by 60,361
  2. 2020Language Models are Few-Shot Learnerscited by 32,419
  3. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  4. 2022Training language models to follow instructions with human feedbackcited by 15,971
  5. 2024The Llama 3 Herd of Modelscited by 14,715
  6. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,599
  7. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  8. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  9. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
  10. 2019Parameter-Efficient Transfer Learning for NLPcited by 4,236
  11. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,181
  12. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,780
38 more
  1. 2022DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generationcited by 3,300
  2. 2023Mistral 7Bcited by 3,164
  3. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,975
  4. 2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,424
  5. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,998
  6. 2023IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Modelscited by 1,464
  7. 2015Microsoft COCO Captions: Data Collection and Evaluation Servercited by 1,462
  8. 2023T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Modelscited by 1,351
  9. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,336
  10. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,302
  11. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,255
  12. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,085
  13. 2023Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formattingcited by 817
  14. 2023Diffusion Model Alignment Using Direct Preference Optimizationcited by 790
  15. 2018Computational Optimal Transportcited by 780
  16. 2023An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuningcited by 639
  17. 2018Texygen: A Benchmarking Platform for Text Generation Modelscited by 491
  18. 2023Directly Fine-Tuning Diffusion Models on Differentiable Rewardscited by 448
  19. 2023Uni-ControlNet: All-in-One Control to Text-to-Image Diffusion Modelscited by 359
  20. 2024SDXL-Lightning: Progressive Adversarial Diffusion Distillationcited by 266
  21. 2023Using Human Feedback to Fine-tune Diffusion Models without Any Reward Modelcited by 252
  22. 2023Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representationscited by 209
  23. 2024ReFT: Representation Finetuning for Language Modelscited by 197
  24. 2023Concept Sliders: LoRA Adaptors for Precise Control in Diffusion Modelscited by 153
  25. 2017Learning to Generate Reviews and Discovering Sentimentcited by 148
  26. 2023Understanding Catastrophic Forgetting in Language Models via Implicit Inferencecited by 144
  27. 2023Navigating Text-To-Image Customization: From LyCORIS Fine-Tuning to Model Evaluationcited by 109
  28. 2024Controllable Generation with Text-to-Image Diffusion Models: A Surveycited by 83
  29. 2024Get What You Want, Not What You Don't: Image Content Suppression for Text-to-Image Diffusion Modelscited by 63
  30. 2023Controlled Text Generation via Language Model Arithmeticcited by 51
  31. 2024Statistical optimal transportcited by 47
  32. 2023SCEdit: Efficient and Controllable Image Diffusion Generation via Skip Connection Editingcited by 37
  33. 2024Whispering Experts: Neural Interventions for Toxicity Mitigation in Language Modelscited by 32
  34. 2024Controlling Large Language Model Agents with Entropic Activation Steeringcited by 29
  35. 2022The Stable Artist: Steering Semantics in Diffusion Latent Spacecited by 23
  36. 2023Steered Diffusion: A Generalized Framework for Plug-and-Play Conditional Image Synthesiscited by 18
  37. 2021Self-conditioning pre-trained language modelscited by 16
  38. 2024Do not think about pink elephant!cited by 4

50 most cited of 51 All 51 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 15; 2026: 53.

68 papers.

Sort:

By area:

  1. 2025A Unified Understanding and Evaluation of Steering Methodscited by 46
  2. 2026Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behaviorcited by 26
  3. 2026ODESteer: A Unified ODE-Based Steering Framework for LLM Alignmentcited by 19
  4. 2026Observing and Controlling Features in Vision-Language-Action Modelscited by 16
  5. 2025Activation Steering with a Feedback Controllercited by 15
  6. 2025LinEAS: End-to-end Learning of Activation Steering with a Distributional Losscited by 14
  7. 2025SAKE: Steering Activations for Knowledge Editingcited by 13
  8. 2025Video Unlearning via Low-Rank Refusal Vectorcited by 13
  9. 2025Steering into New Embedding Spaces: Analyzing Cross-Lingual Alignment Induced by Model Interventions in Multilingual Language Modelscited by 11
  10. 2026Local Linearity of LLMs Enables Activation Steering via Model-Based Linear Optimal Controlcited by 9
  11. 2025SteerVLM: Robust Model Control through Lightweight Activation Steering for Vision Language Modelscited by 8
  12. 2026COLD-Steer: Steering Large Language Models via In-Context One-step Learning Dynamicscited by 8
38 more
  1. 2025Concept Steerers: Leveraging K-Sparse Autoencoders for Test-Time Controllable Generationscited by 7
  2. 2026Sparse Autoencoders are Capable LLM Jailbreak Mitigatorscited by 6
  3. 2026Concept Heterogeneity-aware Representation Steeringcited by 6
  4. 2026Beyond Steering Vector: Flow-based Activation Steering for Inference-Time Interventioncited by 6
  5. 2025Neurons Speak in Ranges: Breaking Free from Discrete Neuronal Attributioncited by 5
  6. 2025DISCO: Disentangled Communication Steering for Large Language Modelscited by 5
  7. 2026Steering at the Source: Style Modulation Heads for Robust Persona Controlcited by 5
  8. 2025Evaluating Sparse Autoencoders for Monosemantic Representationcited by 4
  9. 2026Steering Large Reasoning Models towards Concise Reasoning via Flow Matchingcited by 4
  10. 2026TADA! Tuning Audio Diffusion Models through Activation Steeringcited by 4
  11. 2025Self-Improving Model Steeringcited by 3
  12. 2025DSO: Direct Steering Optimization for Bias Mitigationcited by 3
  13. 2026Conditioned Activation Transport for T2I Safety Steeringcited by 3
  14. 2026Where to Steer: Input-Dependent Layer Selection for Steering Improves LLM Alignmentcited by 3
  15. 2026CausalDetox: Causal Head Selection and Intervention for Language Model Detoxificationcited by 3
  16. 2026Don't Lose Focus: Activation Steering via Key-Orthogonal Projectionscited by 3
  17. 2026Activation Steering of Video Generation Models via Reduced-Order Linear Optimal Controlcited by 3
  18. 2026Beyond Linear Activation Steering: Invertible Latent Transformations for Controlling LLM Behaviorcited by 3
  19. 2026A Geometric Perspective on Composable Emotion Steering in Text-to-Speech Modelscited by 3
  20. 2026Steering Robustness into World Action Models via Mechanistic Interpretability and Optimal Controlcited by 3
  21. 2026GenCtrl -- A Formal Controllability Toolkit for Generative Modelscited by 2
  22. 2026Exploring Diverse Generation Paths via Inference-time Stiefel Activation Steeringcited by 2
  23. 2026ELROND: Exploring and decomposing intrinsic capabilities of diffusion modelscited by 2
  24. 2026Do Models Share Safety Representations? Cross-Model Steering for Safe Visual Generationcited by 2
  25. 2026Conditional Optimal Bridge for Riemannian Activation Steeringcited by 2
  26. 2026Locating and Controlling Implicit Personalization in Large Language Modelscited by 2
  27. 2026Efficient Refusal Ablation in LLM through Optimal Transportcited by 1
  28. 2026SHIFT: Steering Hidden Intermediates in Flow Transformerscited by 1
  29. 2026On The Effectiveness-Fluency Trade-Off In LLM Conditioning: A Systematic Studycited by 1
  30. 2026High-Dimensional Random Projection for Activation Steering in Language Modelscited by 1
  31. 2026DiMaS: Distribution Matching for Steering Vision-Language-Action Modelscited by 1
  32. 2026Semantic Steering for Controllable Generation: Tuning-Free Concept Erasure in Multimodal Diffusion Transformerscited by 1
  33. 2026Reference-Grafting Matches Fine-Tuning at Eliciting Sandbagged Capabilitiescited by 1
  34. 2026A Causal Model for Locating and Unlocking Sandbagging in Model Organismscited by 1
  35. 2026GAPS: Dimension-Level Gates for Conditional Activation Steeringcited by 1
  36. 2025ExpertLens: Activation steering features are highly interpretablecited by 0
  37. 2025Activation Transport Operatorscited by 0
  38. 2026HyperTransport: Amortized Conditioning of T2I Generative Modelscited by 0

50 most cited of 68 All 68 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2024Multi-property Steering of Large Language Models with Dynamic Activation Composition
  2. 2025Steering When Necessary: Flexible Steering Large Language Models with Backtracking
  3. 2025Concept Reachability in Diffusion Models: Beyond Dataset Constraints
  4. 2025Activation Steering for Masked Diffusion Language Models
  5. 2023Semantic Guidance Tuning for Text-To-Image Diffusion Models
  6. 2026NEGATE: Constrained Semantic Guidance for Linguistic Negation in Text-to-Video Diffusion
  7. 2025In-Distribution Steering: Balancing Control and Coherence in Language Model Generation
  8. 2026UniSteer: Text-Guided Flow Matching in Activation Space for Versatile LLM Steering