Channels4World releases 100,000 outdoor RF scenes for wireless models
Original title: CHANNELS4WORLD: An Environment-Scale Dataset for Wireless Foundation Models
The 29.14-TB corpus spans 3.5 and 7 GHz with a 90K/10K scene-level train-test split.
Channels4World is a scene-centered dataset of 100,000 distinct outdoor RF scenes plus a procedural generator for wireless foundation-model pretraining. Each scene links geometry and RF materials to 32 base-station positions, 16,384 UE positions, and multipath records at 3.5 and 7 GHz. The corpus has 52.43 billion physical BS-UE pairs and a 29.14-TB payload. Broader scene collections improved held-out performance under matched training exposure, though scene count alone did not explain the gain.
- Public release includes predefined subsets and separate access to corpus components.
- A deterministic 90K/10K split keeps both frequencies and all scene components together.
- Diffraction materially affects retained-link availability and multipath structure.
- The retention policy cuts storage with little loss in represented aggregate power.
From the original article
1 min read at arxiv.orgWireless foundation models learn reusable representations across tasks, configurations, and deployment environments. Their pretraining requires environmental breadth, yet wireless datasets often sample few sites densely. We introduce Channels4World, a scene-centered dataset of 100,000 distinct outdoor RF scenes and a procedural method for generating them.
By Joonkyu Jang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.