---
format: "aidr-story-markdown/v1"
id: "8964256bb09caf1f152eac5170474cfdef23f7655183ebb7f90b237dc55561ce"
canonical_url: "https://aidr.today/8964256b?lang=vi"
title: "Build Training-Ready Physical AI Datasets with Two Prompts in NVIDIA Cosmos"
lang: "en"
requested_lang: "vi"
available_langs: ["en"]
translation_fallback: "en"
fallback_fields: ["title","summary"]
published_at: "2026-11-18T17:00:00.000Z"
category: "Data"
topics: ["nvidia","agent","data-engineering"]
source_urls: ["https://developer.nvidia.com/blog/build-training-ready-physical-ai-datasets-with-two-prompts-in-nvidia-cosmos/"]
summary: "World model accuracy starts with high-quality, diverse training datasets. Yet capturing rare defects, nighttime incidents, and other challenging scenarios is costly and time-consuming."
---

# Build Training\-Ready Physical AI Datasets with Two Prompts in NVIDIA Cosmos

> [Open the canonical story](<https://aidr.today/8964256b?lang=vi>)

**Published:** 2026-11-18T17:00:00.000Z
**Category:** Data
**Topics:** nvidia, agent, data\-engineering

_English fallback used for: title, summary._

## Summary

World model accuracy starts with high\-quality, diverse training datasets\. Yet capturing rare defects, nighttime incidents, and other challenging scenarios is costly and time\-consuming\.

## Sources

- [Story source](<https://developer.nvidia.com/blog/build-training-ready-physical-ai-datasets-with-two-prompts-in-nvidia-cosmos/>)

