AINeutralarXiv – CS AI · 9h ago6/10
🧠
Mixture of Masters: Sparse Chess Language Models with Player Routing
Researchers introduce Mixture-of-Masters (MoM), a sparse mixture-of-experts chess language model that routes moves through specialized GPT experts trained on individual grandmaster playing styles. The system outperforms dense transformer baselines and maintains interpretability by dynamically selecting which grandmaster persona to channel based on game state.