V2 — Cross-section / topology
2D-view → 3D inference · cannot be shortcut by code · geometric
QA Tasks
Question-answering task families generated by the BabyVision-v2 deterministic pipeline. Each family targets a distinct reasoning skill; every task ships with executable ground truth (no LLM-as-judge).
V3 — Segmentation / referring
pixel-level GT · SSIM / LPIPS · defends against LLM-as-judge
V4 — Manipulation / counterfactual
rotate · slice · deform · geometry-grounded GT
V6 — Multi-step reasoning
multi-tool trace · scored on answer + trace fidelity
Interactive Physics
force control · MPM fracture · MPM granular · FEM cloth · P1/P2 — discover by acting on objects
Spatial Vision
M3D · IXI brain · distance / angle / topology / volume
Live task-instance data lands when QA-Gen artifacts begin shipping. See task-families.md for the canonical definitions.