ARTFEED — Contemporary Art Intelligence

CMU-Drive and V2V-VLA: New Benchmark and Model for Cooperative Autonomous Driving

ai-technology · 2026-08-11

A recent paper on arXiv presents Cooperative Multi-agent Unified Driving with Reasoning (CMU-Drive), a comprehensive closed-loop benchmark designed to assess cooperative autonomous driving involving multiple connected autonomous vehicles (CAVs) in critical safety situations with surrounding traffic. Additionally, the authors introduce Vehicle-to-Vehicle Vision-Language-Action (V2V-VLA), a model that merges cooperative driving into a unified forward pass, simultaneously producing driving actions, future waypoints, language reasoning, and communication strategies. The experiments conducted on CMU-Drive establish the initial benchmark and baseline for cooperative VLA driving. This paper can be found on arXiv under the identifier 2608.07621.

Key facts

  • CMU-Drive is a closed-loop end-to-end benchmark for cooperative autonomous driving.
  • It evaluates multiple connected autonomous vehicles (CAVs) in safety-critical scenarios with background traffic.
  • V2V-VLA is a cooperative Vision-Language-Action model.
  • V2V-VLA integrates cooperative driving into a single forward pass.
  • It jointly generates driving actions, future waypoints, language reasoning, and communication policies.
  • Experiments on CMU-Drive establish the first benchmark and baseline for cooperative VLA driving.
  • The paper is published on arXiv with identifier 2608.07621.
  • The announcement type is new.

Entities

Institutions

  • arXiv

Sources