Large Language Models
23 views
Jailbreaking
Quick Definition
Circumventing AI safety restrictions through prompt techniques
Full Definition
Circumventing AI safety guardrails and content restrictions through clever prompt engineering.
Examples
DAN prompts, roleplay exploits, instruction override
Related Terms
prompt-injection
ai-safety
guardrails-llm
More Large Language Models Terms
Model Merging
Combining fine-tuned LLMs without additional training
Flash Attention
Memory-efficient attention using GPU SRAM block computation
Tool Use
Capability of LLMs to call external tools and APIs
Grounding LLM
Connecting LLM outputs to verifiable external sources
Scaling Laws
Relationships between model size, data, compute, and performance
RoPE
Position encoding using rotation matrices for relative positions