AI Briefing
KOSign in

Hugging Face Introduces Decision Models for Probabilistic, Non-Generative AI Outputs

·2026.11.10 09:00

Key point

The new interface supports choice, score, and noul primitives with a unified API across transformers and llama.cpp.

1 / 2

Details

  • Hugging Face introduces Decision Models, which return probability distributions over predefined answers instead of generating text, avoiding the schema compliance issues and computational overhead of prompting LLMs for JSON.
  • The interface defines three question primitives: Choice (probability distribution over options), Score (expectation value on an ordered scale), and Noul (Bernoulli probability for binary yes/no outcomes).
  • A unified /v1/systemone endpoint is supported by both transformers and llama.cpp, allowing models like autotrust/GEV-26B-Decide and ggml-org/Kev-4B-GGUF to serve decisions without model-specific adaptation.
  • Training is supported via the Trainer API; for example, choice questions are trained with a plain cross-entropy loss over the options, and the Hub now uses a decision-model tag for discovery across different architectures.
ai-infrastructurellama.cpphugging-facetransformersdecision-modelsprobabilistic-ai
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

PyTorchKRHugging Face Publishes Draft Standard for Transformers Model Integration ContractsHuggingFace BlogHugging Face Introduces Model Integration Contracts to Catch LLM Regressions EarlyHuggingFace BlogHugging Face and CMU Launch Project ProtoStar to Make Publicly Funded Science AI-ReadyGeekNewsGoogle acquires Crusoe's Wyoming data center siteRedditNex-AGI Releases Nex-N2.6: 321B Multimodal Agentic Model for Computer UseHacker NewsNvidia Used Circular Financing to Keep GPU Prices High, WSJ Report Claims