Hugging Face adds end-to-end support for Decision Models
ariG23498 · x · 2026-11-11
Hugging Face announces end-to-end support for Decision Models — models that directly output a probability distribution over choices instead of generating text. Support spans transformers (a 'decision' pipeline in Python and CLI), llama.cpp for local inference, and trl/unsloth for fine-tuning. A customer-support routing example shows the model returning department, urgency and refund decisions directly, skipping JSON generation and parsing.
More from coding & agent
- PHPcon Niigata talk: 7 keys to team development with generative AI — 4310sy · 2026-11-14
- Agent skills are everywhere — but they're just markdown files installed via npx — kylegawley · 2026-11-14
- Put your codebase on intermittent fasting: schedule what your agent may touch — vincent_koc · 2026-11-14
- HeyGen open-sources Hyperframes: HTML/CSS animations in, deterministic MP4 out for coding agents — victor_explore · 2026-11-14
- Dynamic tool discovery instead of loading every MCP schema upfront — worth the overhead? — Beginning-Fudge-5025 · 2026-11-14
- Google Antigravity adds orchestrator mode for multi-model coordination in long agentic tasks — Saboo_Shubham_ · 2026-11-14