ICML 2026: AutoMoT reaches SOTA on Bench2Drive and nuScenes

ICML2026 | AutoMoT : B2D & nuScense双SOTA ，重新思考VLM和端到端驾驶的结合

NTU AutoMan Lab, Harvard, and Xiaomi Auto proposed AutoMoT, a unified VLA driving model using a 4B Qwen3-VL Understanding Expert and a 1.6B Action Expert with asynchronous inference, reaching 89.42 DS and 74.09% SR on Bench2Drive with AutoMoT+.

Why it matters: HKR-H/K pass via the async VLM-driving setup and concrete Bench2Drive numbers. The autonomy focus narrows HKR-R, so this sits at the featured threshold rather than the 78+ research tier.