Medical Reasoning
Synthetic evaluation data designed for high-difficulty Chinese medical reasoning: clinical contexts, evidence appraisal and decisions under multiple constraints.
Discover datasets
PCB Assets
Inspect circuit context, schematics, PCB layout and design rules in editable projects.
3D Assets
Inspect editable object geometry, parameters and modeling source.
Foam Tower
Inspect synchronized observations, joint targets and contact traces during grasping, release and stacking.
Handwriting
Inspect the alignment of ordered strokes, simulated writing images, measured joint states and planned targets.
EmbodiedTrace
Inspect robot observations alongside generated subgoals, rationales and structured command labels.
Long-Horizon Terminal Tasks
Turn multi-step reasoning into executable terminal tasks: code repair, scientific analysis and engineering constraints, with environments, reference implementations and separate verifiers.