Generality

What it means

Generality is our attempt to answer: “Is this term a foundational idea that many other concepts depend on, or a more specific technique / variant / implementation detail?”

Higher generality corresponds to broader, more foundational concepts (e.g. “Machine Learning”, “Neural Networks”). Lower generality corresponds to narrower or more specialized concepts (e.g. fine‑tuning methods, particular architecture variants).

How we score it

Each term is scored from its title and summary. We ask a model for 7 independent scores on a 0–1 scale (with some randomness enabled), then:

• Clamp each score to [0, 1]
• Convert each score to an integer 0–1000
• Average the 7 clamped scores and convert that average to 0–1000

The final stored value is that 0–1000 average. (We keep the 7 raw draws separately so we can see variance.)

How to use it

Generality is best used for ranking and sorting (“show me the most foundational concepts first”), not as a precise scientific measurement. Two close scores can be effectively tied.

Caveats

• Model judgment: the score reflects a model’s conceptual judgment from a short summary, not a full literature review.
• Some variance is intentional: taking 7 draws and averaging makes the score more stable than a single draw.
• Scope matters: “general” here means general within AI/ML concepts, not across all of science.

Back to Vocab