Native Robot Reasoning - RoboLab-120 - All: leaderboard

120 tasks (64 simple, 39 moderate, 17 complex), 10episodes/task. Author-run Qwen3.8-27B baseline; other robot-policy values are cited existing leaderboard results; Ours is a distinct trained model, not the base model. Metric: Success rate (%). Source: https://arxiv.org/html/2611.18562v1

Metric: Success rate (%). Source: arxiv.org. 1 model tracked. Updated 17 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Qwen3.8-27BxHigh5.4

Interactive version: theaggregate.ai/benchmark?slug=native-robot-reasoning-robolab-120-all · How It Works · Data refreshed daily, snapshot 2026-11-17.