SWE-Game Benchmark

What is SWE-Game?

We introduce SWE-Game, a benchmark of 247 tasks grounded in 41 executable reference Godot games spanning 13 gameplay categories in 2D and 3D.

Released

2026-09-27

Evaluates

General AI, Geometric reasoning, Vision & 3D, cs.AI

Openness

unknown

Importer

Claire Radar

Review status

rule-auto-admitted

Reported scores

0

Source provenance

• Original evidence https://arxiv.org/abs/2609.33678

SWE-Game paper

• Paper https://arxiv.org/abs/2609.33678

Which sources cite SWE-Game?

• SWE-Game

Related General AI, Geometric reasoning, Vision & 3D, cs.AI benchmarks

• $\tau^\tau$-Bench: An Environment for End-To-End, Realistic Agent Construction
• 2d-fet-bench
• 3DCodeBench
• 5-Dialects-BN
• A Real-World CAD Segmentation Benchmark
• A3S-Bench
• AARRI-Bench
• ABACUS