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

Leduc Hold’em

Benchmarks

Task NameDataset NameSOTA ResultTrend
Opponent ExploitationLeduc Hold'em 3,000 hands × 3 trials, paired seeds (Held-out)
Gain0.821
14
Poker GameplayLeduc Hold'em (test)
NFSP Score41.5
8
Reasoning EvaluationLeduc Hold’em
Hit Rate (HR)2
6
Multi-agent policy generationLeduc hold'em repeated
Population Return49.3
5
Poker Gameplay PerformanceLeduc Hold'em
NFSP24.5
5
Poker Gameplay Performance3-player Leduc Hold'em
Gameplay Performance Score30.8
3
Reasoning Quality Evaluation3-player Leduc Hold'em (test)
Hit Rate (HR)193
3
Strategic game playingLeduc Hold'em held-out (test)
Win Rate0.5389
2
Showing 8 of 8 rows