Semantics as a Shield: Label Disguise Defense (LDD) against Prompt Injection in LLM Sentiment Classification

arXiv:2511.21752 · cs.CL, cs.AI · Submitted 2025-11-23 · Read on arXiv

cs.CL, cs.AI

Submitted: 2025-11-23

Updated: 2026-10-13

Code: https://github.com/Squirrel-333/LDD-prompt-injection-figures

Terminology

• Accuracy — 2× in this paper · explained in 1 episode
• Label Disguise Defense — 2× in this paper
• Few-Shot Learning — 1× in this paper · explained in 4 episodes
• Model-Agnostic — 1× in this paper · explained in 2 episodes
• Prompt Injection Attacks — 1× in this paper · explained in 2 episodes
• Robustness — 1× in this paper · explained in 5 episodes

Sources

• Embedding-based classifiers can detect prompt injection attacks
• ASMR: Adaptive Skeleton-Mesh Rigging and Skinning via 2D Generative Prior
• Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity
• Attention Tracker: Detecting Prompt Injection Attacks in LLMs
• Mistral 7B
• e-GPU: An Open-Source and Configurable RISC-V Graphic Processing Unit for TinyAI Applications
• HyperMagNet: A Magnetic Laplacian based Hypergraph Neural Network
• Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?
• Textbooks Are All You Need
• PromptArmor: Simple yet Effective Prompt Injection Defenses
• Wing Optimisation for a tractor propeller driven Micro Aerial Vehicle
• LLaMA: Open and Efficient Foundation Language Models
• The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions

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