Decision Models in Hugging Face
Web page · https://huggingface.co/blog/decision
First seen by Oriel on .
AI summary
Hugging Face presents a shared interface for decision models, which evaluate predefined questions and return typed decisions with probabilities instead of generating text.
- Requests pair a state with questions whose answers use one of three primitives: choice among options, score on an ordered scale, or noul for a yes/no probability.
- Choice returns probabilities over runtime-defined options; score returns the expected value of an ordered distribution, while noul returns the probability that a statement is true.
- Traditional classifiers have fixed output labels, while zero-shot classifiers accept labels at inference; decision models also support multiple typed questions in one request.
- The article argues that direct decisions avoid generating unnecessary response text and reduce schema-following problems, though constrained decoding can enforce output structure with added computational work.
- Hugging Face aims to standardize the interface while allowing different model implementations, so tools such as llama.cpp, the Hub, and training systems can support them consistently.
#decision models #hugging face #transformers #llm inference
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Blog | https://huggingface.co/blog/decision | https://huggingface.co/blog/decision Read discussion |