Seed
A number that fixes the randomness, so a “random” run can be repeated exactly.
Think of it like
Shuffling a deck the exact same way twice by starting from the same setup — random-looking, but repeatable.
Example
Set the same seed and temperature, and a model can reproduce the identical “random” answer on demand.
How it actually works
Random number generators are actually deterministic given a starting seed. Fixing the seed makes stochastic sampling reproducible, which is invaluable for debugging and fair comparisons. It only helps if the rest of the pipeline is pinned too — change the model or hardware and the guarantee weakens.
For product teams
The knob that makes “random” outputs reproducible for testing and A/B comparisons.
For engineers
The RNG seed initializing pseudo-random sampling; fixes decode randomness for reproducibility.
Related
- Stochastic — Tames stochastic sampling.
- Determinism — Supports determinism.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome