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

FrozenLake

Benchmarks

Task NameDataset NameSOTA ResultTrend
Grid-world NavigationFrozenLake reward reversal Hidden drift
Score100
45
Grid World NavigationFrozenLake (Source)
Score67.5
36
Gridworld NavigationFrozenLake Source
Success Rate9,000
27
Visual PlanningFrozenLake
Success Rate (Level 3)100
19
Grid-World NavigationFrozenLake Drift II
Success Rate8,500
18
Grid NavigationFrozenLake v1.0 (Drift I)
Success Rate7,500
18
Gridworld NavigationFrozenLake Drift II - Topology Shift
Success Rate7,500
18
Gridworld NavigationFrozenLake Drift I - Topology Shift
Success Rate8,500
18
Grid-world NavigationFrozenLake reward reversal Source
Score90
18
Puzzle ReasoningFrozenLake
Success Rate82
17
Agent TaskFrozenLake
Success Rate100
17
Agent Behavior AdaptationFrozenLake (FL) (test)
Loop Ratio0
17
Goal PredictionFrozenLake OOD
Classification Accuracy90
15
Goal PredictionFrozenLake ID
Classification Accuracy100
15
2D discrete grid-world planningFrozenLake
Success Rate83
15
Grid-world NavigationFrozenLake Implicit Hidden Drift
Success Rate (Source)88.8
14
Grid-world NavigationFrozenLake Explicit Structural Drift II
Success Rate (Source)83.7
14
NavigationFrozenLake
Success Rate27
12
Reinforcement LearningFrozenLake
Reward0.94
12
2D NavigationFrozenLake OOD
Success Rate98
11
Spatial decision-makingFrozenLake size 16
Success Rate49.38
11
Spatial decision-makingFrozenLake size 12
Success Rate64.72
11
Spatial decision-makingFrozenLake size 8
Success Rate81.74
11
Spatial decision-makingFrozenLake size 4
Success Rate87.94
11
Multi-turn RL navigationFrozenLake held-out (val)
Success Rate60.7
10
Showing 25 of 54 rows