AUGMANITAI: 1,000+ Terms for Human-LLM Interaction Phenomena
The article introduces AUGMANITAI, an open-access compendium of over 1,000 standardized terms for the various phenomena that emerge when humans interact with large language models (LLMs).
Why it matters
Standardizing terminology for human-LLM interaction patterns is essential for advancing AI research, development, and education.
Key Points
- 1AUGMANITAI provides terminology for patterns in LLM output quality, interaction dynamics, user behavior, prompt engineering, safety/alignment, and evaluation
- 2The compendium follows terminology science principles to enable consistent vocabulary for researchers, developers, and educators working with LLMs
- 3Naming these phenomena is important as many currently lack standardized terminology, hindering communication and understanding
Details
The article explains that while some human-LLM interaction patterns like 'hallucination' and 'sycophancy' have established names, the majority of these phenomena currently lack standardized terminology. AUGMANITAI aims to address this gap by providing a comprehensive compendium of over 1,000 terms covering a wide range of domains, from output quality and interaction dynamics to prompt engineering and safety/alignment. The compendium follows terminology science principles to ensure consistent, precise designations that can be used by researchers, developers, and educators working with large language models. Having a shared vocabulary is crucial for advancing the field and building a better understanding of how humans and AI systems can effectively collaborate.
No comments yet
Be the first to comment