Skin-Deep: A Geometric Diagnostic for Alignment Fragility in Large Language Model Representations

arXiv:2606.22676 · cs.AI · Submitted 2026-06-21 · Read on arXiv

cs.AI

Submitted: 2026-06-21

Updated: 2026-10-01

Code: https://github.com/js-lee-AI/skin-deep

Terminology

• Fine-Tuning — 5× in this paper · explained in 5 episodes
• Alignment — 4× in this paper · explained in 1 episode
• Ablation — 1× in this paper · explained in 1 episode
• Geometric Fragility Score — 1× in this paper
• LoRA Fine-Tuning — 1× in this paper · explained in 3 episodes

Sources

• Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
• The Llama 3 Herd of Models
• Gemma 2: Improving Open Language Models at a Practical Size
• ORPO: Monolithic Preference Optimization without Reference Model
• Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations
• Mistral 7B
• DART: Draft-Agreement Routing for Training-Free Adaptive Thinking Budgets in Hybrid Reasoning Models
• Where Activation Sparsity and KV-Cache Sparsity Cross in LLM Decoding
• Beyond Penalizing Mistakes: Stabilizing Efficiency Training in Large Reasoning Models via Adaptive Correct-Only Rewards
• Answer-Conditioned Chains of Thought Degrade Verifiable-Reasoning Distillation in Large Language Models
• To Isolate or to Score? Model-Adaptive Assessment for Cost-Efficient Multi-Agent RAG
• The Hallucination Signal Is a Mean Shift: Why Simple Probes Suffice
• There is no rose without a thorn: Finding weaknesses on BlenderBot 2.0 in terms of Model, Data and User-Centric Approach
• Steering Llama 2 via Contrastive Activation Addition
• Qwen2.5 Technical Report
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model
• OpenChat: Advancing Open-source Language Models with Mixed-Quality Data
• Shadow Alignment: The Ease of Subverting Safely-Aligned Language Models
• Representation Engineering: A Top-Down Approach to AI Transparency
• Universal and Transferable Adversarial Attacks on Aligned Language Models

Related papers

• 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
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems