No article text was recorded for this version.
contrastive activation analysis (contrastive activation extraction)
Version 1 of 1157 papers158 conceptsfrozen 2026-10-04
Papers in this version
- Blog: Breaking the Mirror: Activation-Based Mitigation of Self-Preference in LLM Evaluators
- Blog: SteeringControl: Holistic Evaluation of Alignment Steering in LLMs
- Blog: Fine-Grained Detoxification via Instance-Level Prefixes for Large Language Models
- Blog: Steering Prosocial AI Agents: Computational Basis of LLM's Decision Making in Social Simulation
- Blog: Internal Activation Revision: Safeguarding Vision Language Models Without Parameter Update
- Blog: L2V-CoT: Cross-Modal Transfer of Chain-of-Thought Reasoning via Latent Intervention
- Blog: Mitigating Overthinking in Large Reasoning Models via Manifold Steering
- Blog: CASteer: Cross-Attention Steering for Controllable Concept Erasure
- Blog: Amortized Latent Steering: Low-Cost Alternative to Test-Time Optimization
- Blog: Refusal Direction is Universal Across Safety-Aligned Languages
- Blog: Activation Steering Decoding: Mitigating Hallucination in Large Vision-Language Models through Bidirectional Hidden State Intervention
- Blog: Component and Dimension Sparsity in Transformer Refusal Mechanisms
- Blog: Understanding (Un)Reliability of Steering Vectors in Language Models
- Blog: Uncovering Latent Chain of Thought Vectors in Large Language Models
- Blog: Sycophancy as compositions of Atomic Psychometric Traits
- Blog: Can Role Vectors Affect LLM Behaviour?
- Blog: Breaking the Mirror: Activation-Based Mitigation of Self-Preference in LLM Evaluators
- Blog: Depth-Wise Activation Steering for Honest Language Models
- Blog: Convergent Linear Representations of Emergent Misalignment
- Blog: Activation Steering for Masked Diffusion Language Models
- Blog: Characterizing and Mitigating Reasoning Drift in Large Language Models
- Blog: Analysing the Generalisation and Reliability of Steering Vectors
- Blog: A gentle push funziona benissimo: making instructed models in Italian via contrastive activation steering
- Blog: AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders
- Blog: A Gentle Push Funziona Benissimo: Making Instructed Models in Italian via Contrastive Activation Steering
- Blog: Improving Instruction-Following in Language Models through Activation Steering
- Blog: Auditing Emotion-Vector-Steered Political Bias in Open-Weight LLMs
- Blog: Benchmarking Mental State Representations in Language Models
- Blog: Debias your Large Multi-Modal Model at Test-Time with Non-Contrastive Visual Attribute Steering
- Blog: PERSONA: Dynamic and Compositional Inference-Time Personality Control via Activation Vector Algebra
- Blog: Understanding Reasoning in Chain-of-Thought from the Hopfieldian View
- Blog: VLM-Guard: Safeguarding Vision-Language Models via Fulfilling Safety Alignment Gap
- Blog: Surgical, Cheap, and Flexible: Mitigating False Refusal in Language Models via Single Vector Ablation
- Blog: The Wedge Questions: Latent Cultural Boundaries in LLMs via Persona Projection Divergence
- Blog: Self-Detoxifying Language Models via Toxification Reversal
- Blog: SCANS: Mitigating the Exaggerated Safety for LLMs via Safety-Conscious Activation Steering
- Blog: Turning the Spell Around: Lightweight Alignment Amplification via Rank-One Safety Injection
- Blog: Do LLMs “Feel”? Emotion Circuits Discovery and Control
- Blog: Steering Language Models with Weight Arithmetic
- Blog: InferAligner: Inference-Time Alignment for Harmlessness through Cross-Model Guidance
- Blog: DESTEIN: Navigating Detoxification of Language Models via Universal Steering Pairs and Head-wise Activation Fusion
- Blog: Nothing in Excess: Mitigating the Exaggerated Safety for LLMs via Safety-Conscious Activation Steering
- Blog: DenseSteer: Steering Small Language Models towards Dense Math Reasoning
- Blog: Icon2: Aligning Large Language Models Using Self-Synthetic Preference Data via Inherent Regulation
- Blog: From Emergence to Control: Probing and Modulating Self-Reflection in Language Models
- Blog: Open the Pandora's Box of LLMs: Jailbreaking LLMs through Representation Engineering
- Blog: Steering Protein Language Models
- Blog: Narrow Finetuning Leaves Clearly Readable Traces in Activation Differences
- Blog: When Reasoning Meets Compression: Understanding the Effects of LLMs Compression on Large Reasoning Models
- Blog: FairSteer: Inference Time Debiasing for LLMs with Dynamic Activation Steering
- Blog: BILLY: Steering Large Language Models via Merging Persona Vectors for Creative Generation
- Blog: Spherical Steering: Geometry-Aware Activation Rotation for Language Models
- Blog: Identifying Cooperative Personalities in Multi-agent Contexts through Personality Steering with Representation Engineering
- Blog: SteerVLM: Robust Model Control through Lightweight Activation Steering for Vision Language Models
- Blog: Balancing Stylization and Truth via Disentangled Representation Steering
- Blog: Steering Llama 2 via Contrastive Activation Addition
- Blog: Analysing the Generalisation and Reliability of Steering Vectors
- Blog: Reducing Hallucinations in Large Vision-Language Models via Latent Space Steering
- Blog: Delta Activations: A Representation for Finetuned Large Language Models
- Blog: Improving Activation Steering in Language Models with Mean-Centring
Showing 60 of 157 papers.
History & citation
Current · Permanent link
- This direction split from Steering Model Behavior with Contrastive Activation Vectors (54% overlap)
Cite this direction
Latest version (a living page):
@misc{lacuna_contrastive_activation_analysis_contrast_42a7f793,
title = {contrastive activation analysis (contrastive activation extraction)},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/contrastive-activation-analysis-contrastive-activation-extraction/rdr_42a7f793c3204d049e4c6fdf13f9b5db},
note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}This exact version (version 1, frozen text):
@misc{lacuna_contrastive_activation_analysis_contrast_42a7f793_v1,
title = {contrastive activation analysis (contrastive activation extraction)},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/contrastive-activation-analysis-contrastive-activation-extraction/rdr_42a7f793c3204d049e4c6fdf13f9b5db/v/1},
note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}