# Dongxi NLP explains SFT, RLAF, and multimodal training for Drex decision models

> Original title: 如何训练一个 SOTA Foundation Decision Model?

- Source: Dongxi NLP
- Published: 2026-11-06T13:39:32.000Z
- Bad Influence: https://badinfluence.news/items/pz0qfo4ps52lpemcudzmgfnrf
- Original: https://x.com/dongxi_nlp/status/2118699806767857744

## Summary

Dongxi NLP posts a recipe for training a SOTA foundation decision model, citing the NaceAI Drex Technical Report. The post says Drex combines supervised fine-tuning, RLAF with KL regularization, and vision fine-tuning for multimodal support. It reports that Drex Max reaches 94.0% accuracy on Decision Bench v4.

## Post

如何训练一个 SOTA Foundation Decision Model?

Drex Technical Report 给出了完整 recipe:

核心是 SFT + RLAF + multimodal training pipeline

- SFT builds foundational decision-making capabilities using labeled and teacher-generated data
- RLAF refines decision probabilities using adjusted feedback with KL regularization
- Multimodal support through vision fine-tuning

Drex Max 在 Decision Bench v4 上达到 94.0% accuracy，展现出接近 SOTA 的结构化决策能力，同时具备较低的 inference latency。

推荐阅读！

## Quoted post @NaceAI

> How are Decision Models Trained?
> 
> Today, we are publishing the Drex Technical Report for the developer community! 🚀
> 
> Discover details on how the State-of-the-art Open-Source decision models are trained with RLAF.
> 
> Build more at: https://console.nace.ai/ (try for free)
> 
> P.S. Drex Max 1.1 and Drex for Finance are closed sourced and designed for enterprise use-cases

https://x.com/NaceAI/status/2118695842551103645