← All benchmarks

EmbodiedLab Process Score (T0)

Agents · 2026-10-26

EmbodiedLab's process score at T0 over the same 30 experiments and runs as the headline success rate. Each run earns 10 for each of four milestones up to sufficient readings (acting on the apparatus, a measurable state, a first valid reading, enough readings), 30 for a result within tolerance with every other gate held and 30 for restoring the bench, so a successful run scores 100 and a failed run at most 70. The score is the mean over each experiment's runs and then over the experiments, on a 0-100 scale.

Top models (higher is better)