Cover Image

Towards Understanding Subliminal Learning: When and How Hidden Biases Transfer

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Schrodi, Simon, Kempf, Elias, Barez, Fazl, Brox, Thomas
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

When and How Does CLIP Enable Domain and Compositional Generalization?
by: Kempf, Elias, et al.
Published: (2025)

Understanding Addition in Transformers
by: Quirke, Philip, et al.
Published: (2023)

SafetyNet: Detecting Harmful Outputs in LLMs by Modeling and Monitoring Deceptive Behaviors
by: Chaudhary, Maheep, et al.
Published: (2025)

Same Question, Different Words: A Latent Adversarial Framework for Prompt Robustness
by: Fu, Tingchen, et al.
Published: (2025)

Towards Interpretable Sequence Continuation: Analyzing Shared Circuits in Large Language Models
by: Lan, Michael, et al.
Published: (2023)

Simple LLM Baselines are Competitive for Model Diffing
by: Kempf, Elias, et al.
Published: (2026)

Visualizing Neural Network Imagination
by: Wichers, Nevan, et al.
Published: (2024)

Interpreting Context Look-ups in Transformers: Investigating Attention-MLP Interactions
by: Neo, Clement, et al.
Published: (2024)

Eureka-Moments in Transformers: Multi-Step Tasks Reveal Softmax Induced Optimization Problems
by: Hoffmann, David T., et al.
Published: (2023)

What Drives Compositional Generalization? The Importance of Continuous Training Objectives in Visual Generative Models
by: Farid, Karim, et al.
Published: (2025)

Subliminal Learning is a LoRA Artifact
by: Nief, Todd, et al.
Published: (2026)

Beyond Linear Steering: Unified Multi-Attribute Control for Language Models
by: Oozeer, Narmeen, et al.
Published: (2025)

Scaling sparse feature circuit finding for in-context learning
by: Kharlapenko, Dmitrii, et al.
Published: (2025)

Emergent and Subliminal Misalignment Through the Lens of Data-Mediated Transfer
by: Askin, Baris, et al.
Published: (2026)

Learning Through Noise: Why Subliminal Learning Works and When It Fails
by: Brockers, Vincent C., et al.
Published: (2026)

Query Circuits: Explaining How Language Models Answer User Prompts
by: Wu, Tung-Yu, et al.
Published: (2025)

Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders
by: Lan, Michael, et al.
Published: (2024)

Is Mamba Capable of In-Context Learning?
by: Grazzi, Riccardo, et al.
Published: (2024)

Subliminal Learning: Language models transmit behavioral traits via hidden signals in data
by: Cloud, Alex, et al.
Published: (2025)

Rethinking Safety in LLM Fine-tuning: An Optimization Perspective
by: Kim, Minseon, et al.
Published: (2025)

Understanding Addition and Subtraction in Transformers
by: Quirke, Philip, et al.
Published: (2024)

Concept Bottleneck Models Without Predefined Concepts
by: Schrodi, Simon, et al.
Published: (2024)

On the Hidden Objective Biases of Group-based Reinforcement Learning
by: Fontana, Aleksandar, et al.
Published: (2026)

Constrained Reinforcement Learning for Safe Heat Pump Control
by: Zhang, Baohe, et al.
Published: (2024)

Learning with Hidden Factorial Structure
by: Arnal, Charles, et al.
Published: (2024)

When Are Learning Biases Equivalent? A Unifying Framework for Fairness, Robustness, and Distribution Shift
by: Mehta, Sushant
Published: (2025)

Sustained Gradient Alignment Mediates Subliminal Learning in a Multi-Step Setting: Evidence from MNIST Auxiliary Logit Distillation Experiment
by: Kitkana, Chayanon, et al.
Published: (2026)

Distributional Preference Learning: Understanding and Accounting for Hidden Context in RLHF
by: Siththaranjan, Anand, et al.
Published: (2023)

Towards Context-Aware Domain Generalization: Understanding the Benefits and Limits of Marginal Transfer Learning
by: Müller, Jens, et al.
Published: (2023)

Constrained Reinforcement Learning with Smoothed Log Barrier Function
by: Zhang, Baohe, et al.
Published: (2024)

Subliminal Effects in Your Data: A General Mechanism via Log-Linearity
by: Aden-Ali, Ishaq, et al.
Published: (2026)

Understanding Transferable Representation Learning and Zero-shot Transfer in CLIP
by: Chen, Zixiang, et al.
Published: (2023)

Two Effects, One Trigger: On the Modality Gap, Object Bias, and Information Imbalance in Contrastive Vision-Language Models
by: Schrodi, Simon, et al.
Published: (2024)

Best-of-N Jailbreaking
by: Hughes, John, et al.
Published: (2024)

Rethinking AI Cultural Alignment
by: Bravansky, Michal, et al.
Published: (2025)

Large Language Models Relearn Removed Concepts
by: Lo, Michelle, et al.
Published: (2024)

Toward a Metrology for Artificial Intelligence: Hidden-Rule Environments and Reinforcement Learning
by: Mathew, Christo, et al.
Published: (2025)

When Privacy Isn't Synthetic: Hidden Data Leakage in Generative AI Models
by: Mustaqim, S. M., et al.
Published: (2025)

Base Models Know How to Reason, Thinking Models Learn When
by: Venhoff, Constantin, et al.
Published: (2025)

Hidden Biases of End-to-End Driving Datasets
by: Zimmerlin, Julian, et al.
Published: (2024)