List your company

Policy Bias Reduction on 3-state × 2-action tabular MDP

4.5Fast-only Biasnaive
3.6769.23814.820.362Jun 25, 2026

Evaluation results

MethodLinks
naiveK=2, E[Delta]=25, rho=12026.064.54.5111
waitK=2, E[Delta]=25, rho=12026.064.50.1727.120.326
Retrace-AK=2, E[Delta]=25, rho=12026.064.53.41.51.31.2
Retroactive Advantage Correction (RAC)K=2, E[Delta]=25, rho=12026.064.50.0947.920.91
Retroactive Advantage Correction (RAC)K=1, sigma_f=0.5, Delta_k={1, ..., 5}2026.068.20.613.712.1—
Retroactive Advantage Correction (RAC)K=2, sigma_f=0.5, Delta_k={1, ..., 5}2026.0614.30.347.944.2—
Retroactive Advantage Correction (RAC)K=3, sigma_f=0.5, Delta_k={1, ..., 5}2026.0618.10.822.720.8—
Retroactive Advantage Correction (RAC)K=4, sigma_f=0.5, Delta_k={1, ..., 5}2026.0621.60.73330.5—
Retroactive Advantage Correction (RAC)K=5, sigma_f=0.5, Delta_k={1, ..., 5}2026.0624.40.639.836.9—
Retroactive Advantage Correction (RAC)K=5 (sat), kernel=saturated row-stochastic (eta=0), sigma_f=0.5, Delta_k={1, ..., 5}2026.0625.10.5248.345—