MazeBench Shows AI Agents Struggle in 3D Puzzles

MazeBench, a free 3D open-world spatial reasoning benchmark, reveals that agents like GPT-6.1 Sol excel at 2D puzzles but struggle in 3D, preferring a top-down view and rarely rotating the camera.

2026-10-02 ~ 2026-10-02 · 2 related posts

• GPT-6.1 Sol Masters 2D Puzzles but Struggles in 3D, Preferring Top-Down Views — patience_cave · 2026-10-02
• MazeBench goes free online: agents prefer top-down views and stall in 3D puzzles — patience_cave · 2026-10-02