Subliminal Transfer of Unsafe Behaviors in AI Agent Distillation
cs.AI
Submitted: 2026-04-16
Updated: 2026-11-15
Terminology
Sources
- On the Opportunities and Risks of Foundation Models
- Black-Box Access is Insufficient for Rigorous AI Audits
- Bias Propagation in Federated Learning
- Diffusion Policy Attacker: Crafting Adversarial Attacks for Diffusion-based Policies
- Subliminal Learning: Language models transmit behavioral traits via hidden signals in data
- Who's Harry Potter? Approximate Unlearning in LLMs
- RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models
- FACET: Teacher-Centred LLM-Based Multi-Agent Systems-Towards Personalized Educational Worksheets
- Alignment faking in large language models
- Unsolved Problems in ML Safety
- Distilling the Knowledge in a Neural Network
- Distilling LLM Agent into Small Models with Retrieval and Code Tools
- Model evaluation for extreme risks
- Evaluating the Social Impact of Generative AI Systems in Systems and Society
- Llama 2: Open Foundation and Fine-Tuned Chat Models
- Jailbroken: How Does LLM Safety Training Fail?
Related papers
- Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Embeddings for Preferences, Not Semantics
- Can We Trust Item Response Theory for AI Evaluation?
- Diagnosing Task Insensitivity in Language Agents