Quantifying Subliminal Behavioral Transfer Ratios in Language Model Distillation
cs.LG, cs.AI, cs.CL
Submitted: 2026-06-09
Updated: 2026-11-11
Terminology
Sources
- SALT: Steering Activations towards Leakage-free Thinking in Chain of Thought
- JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models
- Beyond Black-Box Obfuscation: Mechanistic Analysis and Defense of White-Box Monitors
- Subliminal Learning: Language models transmit behavioral traits via hidden signals in data
- Distilling the Knowledge in a Neural Network
- Mechanistic Anomaly Detection for "Quirky" Language Models
- Playing Devil's Advocate: Off-the-Shelf Persona Vectors Rival Targeted Steering for Sycophancy
- A Comprehensive Survey on Knowledge Distillation
- Mechanistic origins of catastrophic forgetting: why RL preserves circuits better than SFT?
- Llama 2: Open Foundation and Fine-Tuned Chat Models
- Steering Language Models With Activation Engineering
- Qwen2.5 Technical Report
- Representation Engineering: A Top-Down Approach to AI Transparency
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis