Homeopathic Learning: Diluting poisoned fine-tuning data can strengthen its effect
cs.AI
Submitted: 2026-11-06
Updated: 2026-11-06
Code: https://github.com/mjclarke94/homeopathic_learning
Terminology
Sources
- Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs
- Subliminal Learning Is Steering Vector Distillation
- Poisoning Web-Scale Training Datasets is Practical
- DeepSeek-V3 Technical Report
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- Phantom Transfer: Data Poisoning can Survive Data-Level Defences
- Conditional misalignment: common interventions can hide emergent misalignment behind contextual triggers
- Beyond Owls: Subliminal Learning Can Transfer Learned Capabilities and Backdoors
- Gemma 3 Technical Report
- You Didn't Have to Say It like That: Subliminal Learning from Faithful Paraphrases
- BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
- LoRA: Low-Rank Adaptation of Large Language Models
- Kimi K3: Open Frontier Intelligence
- Platypus: Quick, Cheap, and Powerful Refinement of LLMs
- Subliminal Learning is a LoRA Artifact
- Qwen3 Technical Report
- Universal Jailbreak Backdoors from Poisoned Human Feedback
- Towards Understanding Subliminal Learning: When and How Hidden Biases Transfer
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples
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