G-11 (object reappearance) has a shape move off-screen in a direction and return along the same path — teaches trajectory and object persistence. G-25 (separate object spinning) is a shape that rotates in place then translates horizontally to a target position — multi-step motion sequencing. G-33 (visual jenga) is a stack of objects that get removed one by one from top to bottom — sequential extraction with implicit physics ordering. O-29 (ballcolor) is ball tracking tasks with color — motion following plus identity preservation. O-52 (traffic light) is discrete state transitions, lights switching on/off between green and gray — teaches the model that state changes are crisp, not gradual. O-75 (communicating vessels) is fluid equalizing between connected tubes based on pressure — continuous physics simulation over time. O-87 (fluid diffusion) is ink spreading in water — another continuous physical transformation but with expansion rather than equalization.G-35 (hit target after bounce) is a ball with an initial direction that bounces off walls following reflection laws to hit a target — pure trajectory prediction with physics constraints. O-30 (bookshelf) is book rearrangement on shelves — the specific task VBVR highlighted where their model beat Sora 2.O-7 (shape color change) is a single transformation — shape changes from one color to another. O-8 (shape rotation) is a shape rotating by a specific angle. O-13 (outline then move) is two sequential steps: change a shape's outline style, then move it to a new position. O-14 (scale then outline) is also two steps: scale a shape up or down, then change its outline. These four together teach the model that instructions are ordered and each step completes before the next begins.G-13 (grid number sequence) is filling in number patterns on a grid. G-17 (grid avoid red block) is pathfinding on a grid while avoiding obstacles. G-31 (directed graph navigation) is finding the shortest path through a directed graph. G-41 (grid highest cost) is evaluating spatial values on a grid to find the optimal path. O-24 (domino chain) is a sequential cascade where dominoes fall until they hit a gap — teaches causal chains and stopping conditions. O-34 (dot to dot) is connecting numbered dots in sequence — ordered drawing. O-47 (sliding puzzle) is tile rearrangement under constraints, like a 15-puzzle. O-83 (planar warp) is warping a grid to align with a target quadrilateral — geometric transformation.O-1 (color mixing) is RGB additive mixing where two light sources combine and the result fills a target zone — rule-based continuous process. O-33 (counting objects) is exactly what it sounds like — count things correctly. G-3 (stable sort) is arranging objects by a rule while preserving relative order. G-37 (symmetry random) is completing a pattern by mirroring across an axis. O-21 (construction blueprint) is fitting a correct puzzle piece into a gap in a structure. G-44 (BFS) is breadth-first search traversal of a graph — systematic layer-by-layer exploration.