Connect Four
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Connect Four
86Best-Response Win Rate
5
Connect Four 100% optimal adversary (test-time)
-0.98Avg Return
3
Connect Four 70% optimal adversary (test)
0.02Average Return
3
Connect Four 50% optimal adversary (test-time)
0.11Average Return
3
Connect Four 30% optimal adversary (test-time)
0.55Average Return
3
Connect Four held-out (test)
21.55Win Rate
2
Connect Four
—Primary metric
0