Correct, Don't Delete: Mitigating Emergent Misalignment with Corrective Supervision

arXiv:2609.37624 · cs.CL, cs.AI, cs.LG · Submitted 2026-09-29 · Read on arXiv

cs.CL, cs.AI, cs.LG

Submitted: 2026-09-29

Updated: 2026-09-29

Terminology

• Fine-Tuning — 14× in this paper · explained in 5 episodes
• Alignment — 10× in this paper · explained in 1 episode
• Calibration — 5× in this paper · explained in 7 episodes
• Influence Functions — 5× in this paper · explained in 2 episodes
• Benchmark — 4× in this paper · explained in 2 episodes
• Accuracy — 3× in this paper · explained in 1 episode
• Generalization — 3× in this paper · explained in 13 episodes
• Reward Hacking — 3× in this paper · explained in 8 episodes
• AI Alignment — 2× in this paper · explained in 2 episodes
• Alignment Scores — 2× in this paper · explained in 1 episode
• In International Conference on Learning Representations — 2× in this paper
• In International Conference on Machine Learning — 2× in this paper
• AUC — 1× in this paper · explained in 2 episodes
• Adam — 1× in this paper · explained in 2 episodes
• Adaptation — 1× in this paper · explained in 1 episode
• GRPO — 1× in this paper · explained in 6 episodes
• Tokenization — 1× in this paper · explained in 2 episodes

Sources

• Persona Vectors: Monitoring and Controlling Character Traits in Language Models
• Conditional misalignment: common interventions can hide emergent misalignment behind contextual triggers
• Gemma 4 Technical Report
• Studying Large Language Model Generalization with Influence Functions
• Measuring Massive Multitask Language Understanding
• LoRA: Low-Rank Adaptation of Large Language Models
• In-Training Defenses against Emergent Misalignment in Language Models
• A Rank Stabilization Scaling Factor for Fine-Tuning with LoRA
• Bayesian Influence Functions for Hessian-Free Data Attribution
• Natural Emergent Misalignment from Reward Hacking in Production RL
• Safety Pretraining: Toward the Next Generation of Safe AI
• TAME: Token Attribution and Masking for Emergent misalignment
• Emergent Misalignment Recruits a Pre-existing Persona Subspace
• How Much of Your Data Can Suck? Thresholds for Domain Performance and Emergent Misalignment in LLMs
• Qwen2.5 Technical Report
• An Emergent Mirage: Is Emergent Misalignment and Realignment Indeed a Robust Phenomenon?
• Convergent Linear Representations of Emergent Misalignment
• Emergent Misalignment is Easy, Narrow Misalignment is Hard
• Self-Recognition Finetuning can Prevent and Reverse Emergent Misalignment
• Inoculation Prompting: Eliciting traits from LLMs during training can suppress them at test-time

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering