TsuGO Benchmark Evaluates LLM Search Efficiency via Go Problems
August 14, 2026
TsuGO uses Go life-and-death problems to measure how LLMs organize search and allocate reasoning resources. The benchmark requires models to perform candidate generation, branch comparison, and backtracking within a closed, verifiable solution space.
HOW THIS AFFECTS YOU
●
researcherYou can use this to move beyond final-answer accuracy and evaluate how models actually plan reasoning paths.