Share your thoughts, 1 month free Claude Pro on usSee more

Language-based Games on HouseBuild (test)

100HP

GRPO

Updated 5mo ago

Evaluation Results

Method	Links
GRPO 2026.01		100	54.6
AC 2026.01		100	55.9
Raw Model 2026.01		99.6	43.2
CoLLM-CC 2026.01		86.4	52.7
MAGRPO 2026.01		80.2	50.9
CoLLM-DC 2026.01		43.8	46.8
Pipeline 2026.01		30.6	41.3
Discussion 2026.01		27.6	38.1
Parallel 2026.01		21.8	46.1