Public articles linked to the same research event.
NVIDIA Technical Blog NVIDIA transferred its general-purpose long-horizon agent architecture AVO from GPU-kernel optimization to the ARC-AGI-3 interactive reasoning benchmark, completing all 25 public-set environments and all 183 levels with a 100.00 RHAE score in 6,624 environment actions, and reports that persistent memory and supervision sustain autonomous progress across domains.
NVIDIA transferred its general-purpose long-horizon agent architecture AVO from GPU-kernel optimization to the ARC-AGI-3 interactive reasoning benchmark, completing all 25 public-set environments and all 183 levels with a 100.00 RHAE score in 6,624 environment actions, and reports that persistent memory and supervision sustain autonomous progress across domains.
NVIDIA transferred its general-purpose long-horizon agent architecture AVO from GPU-kernel optimization to the ARC-AGI-3 interactive reasoning benchmark, completing all 25 public-set environments and all 183 levels with a 100.00 RHAE score in 6,624 environment actions, and reports that persistent memory and supervision sustain autonomous progress across domains.
NVIDIA transferred its general-purpose long-horizon agent architecture AVO from GPU-kernel optimization to the ARC-AGI-3 interactive reasoning benchmark, completing all 25 public-set environments and all 183 levels with a 100.00 RHAE score in 6,624 environment actions, and reports that persistent memory and supervision sustain autonomous progress across domains.