Score: 1

Improving Robustness of AlphaZero Algorithms to Test-Time Environment Changes

Published: September 4, 2025 | arXiv ID: 2509.04317v1

By: Isidoro Tamassia, Wendelin Böhmer

Potential Business Impact:

Makes smart game players adapt to new rules.

Business Areas:
A/B Testing Data and Analytics

The AlphaZero framework provides a standard way of combining Monte Carlo planning with prior knowledge provided by a previously trained policy-value neural network. AlphaZero usually assumes that the environment on which the neural network was trained will not change at test time, which constrains its applicability. In this paper, we analyze the problem of deploying AlphaZero agents in potentially changed test environments and demonstrate how the combination of simple modifications to the standard framework can significantly boost performance, even in settings with a low planning budget available. The code is publicly available on GitHub.

Country of Origin
🇧🇪 Belgium

Repos / Data Links

Page Count
20 pages

Category
Computer Science:
Artificial Intelligence