A number that initializes the random component of a model's sampling. Pinning the seed (with a low temperature and fixed model version) makes outputs reproducible across runs.
Definition: A number that initializes the random component of a model's sampling. Pinning the seed (with a low temperature and fixed model version) makes outputs reproducible across runs.
A number that initializes the random component of model sampling. Same seed, same prompt, same model version, temperature near zero: same output, mostly. Most consumer chat interfaces hide seed control. APIs expose it. Reproducibility starts here.
Model parameter controlling how random the output is during sampling. Temperature 0 makes the model pick the most probable next token at each step; higher values introduce more variation. Pin to 0 for reproducibility, raise for exploration.
The text input you send to a language model. Most 'AI doesn't work' complaints trace back to prompt quality before they trace to model quality.
The structured way one piece of software talks to another. API access to a foundational model gives you control over prompts, parameters, version pinning, and data flow that no chat interface offers.