BabyVision-v2 / QA Tasks
loading…

QA Tasks

Public demo cases spanning the scientific domains BabyVision-v2 covers. Each card opens a live, interactive 3D case; every task ships with executable ground truth (no LLM-as-judge).

Biology

proteins & small molecules · RCSB PDB · AlphaFold DB · PubChem 3D conformers

View demo case →

Medical

CT / MRI / NIfTI / DICOM / NRRD · Decathlon hippocampus · IXI brain · M3D · Slicer NRRD

View demo case →

机械 · Mechanical / CAD

glTF · STEP / IGES B-rep · STL / OBJ · USD · SOLIDWORKS · marble warehouse

View demo case →

建筑 · Architecture

IFC/BIM glTF · structural frame · bridge LiDAR point cloud · per-point structural classes

View demo case →

Embodied Scene

Genesis / Jenga simulation · physics sandbox · animated GLB

View demo case →

Astronomy

solar-system ephemeris · planet / ring geometry · exoplanet elements

View demo case →

Expert Seed Cases

expert-v2 · reviewed

Hand-crafted seed cases contributed by domain experts, spanning three domains — Chemistry, Biology, and Medical. Each is reasoning-heavy and solvable from the 3D geometry alone (no metadata lookup or single-call solver). Open a case to inspect the exact structure in a live interactive viewer, alongside the expert review (tier, reviewer comments, and the V1→V2 answer diff).

loading…

The full generated QA set is listed below and grows as QA-Gen ships more artifacts. See task-families.md for the canonical definitions.