NVIDIA Ising Calibration 1.5 Automates Qubit Tuning

Ising Calibration 1.5 targets one of quantum computing’s least glamorous but most consequential bottlenecks: turning noisy measurement data into reliable calibration decisions. Its developer significance lies less in replacing quantum engineers than in making calibration workflows programmable and repeatable.

• –Built on Gemma 4 31B, with image-and-text inputs for quantum experiment analysis
• –Covers six tasks, including plot interpretation, parameter extraction, diagnostic reasoning, and success classification
• –QCalEval results show stronger zero-shot performance than the base Gemma model, while few-shot performance remains competitive
• –NVIDIA provides an agent blueprint, NIM container, hosted endpoint, and benchmark tooling for integration
• –Outputs still require domain-expert validation before controlling real quantum hardware