logits
The raw, unnormalised scores a model assigns to every token in its vocabulary as candidates for the next position.
In more detail
Softmax turns logits into probabilities that sum to one, after temperature has divided them. Some APIs expose the resulting log probabilities of the chosen tokens, which helps measure confidence and debug classifications.
Related terms
- softmax — The function that turns a list of scores into probabilities by exponentiating each one and dividing by the total.
- temperature — The sampling setting that divides the logits before softmax: values below one sharpen the distribution towards the likeliest token, values above one flatten it.
- sampling (decoding) — Choosing each next token at random from the model's probability distribution, shaped by settings such as temperature, top k and top p.
Taught in
- How The Next Token Is Chosen — AI Engineering: LLM Apps, RAG & Agents
- Keyword And Hybrid Search — AI Engineering: LLM Apps, RAG & Agents
References
- Holtzman et al. (2019): The Curious Case of Neural Text Degeneration (nucleus sampling)
- OpenAI: Chat Completions API reference (temperature, penalties, logprobs, stop)