English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Large-scale AI systems that combine search and learning have reachedsuper-human levels of performance in game-playing, but have also been shown tofail in surprising ways. The brittleness of such models limits their efficacyand trustworthiness in real-world deployments. In this work, we systematicallystudy one such algorithm, AlphaZero, and identify two phenomena related to thenature of exploration. First, we find evidence of policy-value misalignment --for many states, AlphaZero's policy and value predictions contradict eachother, revealing a tension between accurate move-selection and value estimationin AlphaZero's objective. Further, we find inconsistency within AlphaZero'svalue function, which causes it to generalize poorly, despite its policyplaying an optimal strategy. From these insights we derive VISA-VIS: a novelmethod that improves policy-value alignment and value robustness in AlphaZero.Experimentally, we show that our method reduces policy-value misalignment by upto 76%, reduces value generalization error by up to 50%, and reduces averagevalue error by up to 55%.

Policy-Value Alignment and Robustness in Search-based Multi-Agent Learning