List your company

ResearchFormerly WizWand

Policy Evaluation

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

PolicyBench Overall Average

66.34Accuracy

11

Apr 15, 2026

PolicyBench Level 3 US

77Accuracy

11

Apr 15, 2026

PolicyBench Level 3 CN

80.34Accuracy

11

Apr 15, 2026

PolicyBench Level 2 (US)

68.95Accuracy

11

Apr 15, 2026

PolicyBench Level 2 (CN)

62.92Accuracy

11

Apr 15, 2026

PolicyBench Level 1 (US)

59.33Accuracy

11

Apr 15, 2026

PolicyBench Level 1 (CN)

62.02Accuracy

11

Apr 15, 2026

Out-of-distribution

96.2Return (r)

8

Jun 18, 2026

In-distribution

98.4r Score

8

Jun 18, 2026

20-Link Pole off-policy

415.37Sum of sqrt MSE

7

Feb 26, 2026

400-State Random MDP on-policy

24.74Sum of sqrt MSE

7

Feb 26, 2026

14-State Boyan Chain on-policy

25.06Sum of sqrt MSE

7

Feb 26, 2026

20-Link Pole off-policy

4.17MSE

7

Feb 26, 2026

20-Link Pole on-policy

4.26MSE

7

Feb 26, 2026

Cart-Pole off-policy perfect features

0.17MSE

7

Feb 26, 2026

Cart-Pole on-policy, perfect features

0.15MSE

7

Feb 26, 2026

Cart-Pole off-policy, impoverished features

2.33MSE

7

Feb 26, 2026

Cart-Pole on-policy, impoverished features

2.37MSE

7

Feb 26, 2026

400-State Random MDP (off-policy)

0.11MSE

7

Feb 26, 2026

400-State Random MDP on-policy

0.07MSE

7

Feb 26, 2026

14-State Boyan Chain on-policy

0.1MSE

7

Feb 26, 2026

BPIC real logs 2017

MDP-based RL method

1,731.108Average KPI (Mπ)

6

Jun 15, 2026

BPIC real logs 2012

MDP-based RL method

374.966Average KPI (Mπ)

6

Jun 15, 2026

2x2 Queueing Network rho = (0.95, 0.85)

1,185.4Expected Value

6

Jun 11, 2026

2x2 Queueing Network rho = (0.90, 0.90)

1,316.9Value

6

Jun 11, 2026