---
title: "CFM-Bench: A Unified Multi-Domain, Multi-Task Benchmark for Channel Foundation Models"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.14975v1"
permalink: "https://feed.astralor.com/items/itm_bef843e461bfc4dbe18c6c5c"
published: "2026-07-16T13:29:02.000Z"
score: 81
channel: "ai"
tags: ["CFM-Bench","Channel Foundation Model","Benchmark","Wireless Communication","AI for Communications"]
---

# CFM-Bench: A Unified Multi-Domain, Multi-Task Benchmark for Channel Foundation Models

## AI commentary

> This paper is crucial for teams working at the intersection of wireless communications and AI. It tackles the industry pain point of unfair model comparisons and lack of standardized tasks and data. CFM-Bench not only provides a standardized pipeline but also strictly regulates pretraining data usage to prevent cheating. If you are researching channel representations or developing communication foundation models, this benchmark will be the go-to tool for validating and comparing your model's performance.

## At a glance

CFM-Bench is a unified benchmark for channel foundation models (CFMs), addressing fragmented evaluation pipelines and unfair comparisons. It curates six channel configurations spanning 3GPP statistical simulation, ray tracing, industrial/aerial measurements, and vehicular multimodal simulation. The benchmark forbids external pretraining corpora and reserves the training split solely for downstream fine-tuning. It organizes six task groups across PHY channel intelligence, RAN decision intelligence, and ISAC. It provides a common substrate for comparing transferability across models, domains, and tasks.

## Key points

- Curates six channel configurations including statistical simulation, ray tracing, real-world measurements, and multimodal simulation
- Forbids external pretraining corpora; training split exclusively for downstream fine-tuning
- Covers PHY, RAN decision, and ISAC application dimensions
- Provides a fair comparison framework for transferability across models, domains, and tasks

## Full text

Channel foundation models (CFMs) are developing rapidly, with recent studies reporting benefits from pretraining across downstream wireless tasks. Yet CFMs are commonly evaluated in model-specific pipelines with different data, radio configurations, partitions, adaptation procedures, task definitions, and metrics. Reported comparisons therefore tend to show that pretraining improves over supervised training from scratch within one pipeline, but neither rank CFMs nor compare them fairly with task-specific models. We release CFM-Bench, a unified multi-domain, multi-task benchmark designed to address this gap. It curates six channel configurations spanning 3GPP statistical simulation, two independent ray-tracing pipelines, industrial and aerial measurements, and synchronized vehicular multimodal simulation. Official partitions isolate complete trajectories, measurement sessions, vehicle links, simulation realizations, or buffered spatial regions. CFM-Bench does not prescribe an external pretraining corpus or strategy; no benchmark split may be used for foundation-model pretraining, and the official training split is reserved exclusively for downstream fine-tuning. The benchmark additionally requires disclosure of all data used during model development and prohibits training-stage use of official test units. Six task groups are organized along three CFM application dimensions: physical-layer (PHY) channel intelligence, radio-access-network (RAN) decision intelligence, and integrated sensing and communication (ISAC). They cover CSI feedback, frequency and temporal channel extrapolation, propagation-state classification, current- and future-beam prediction, and single-frame and temporal localization. CFM-Bench provides a common substrate for comparing the transferability of channel representations across models, domains, and tasks.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_bef843e461bfc4dbe18c6c5c