pages/index.md

Repro - NonZero: Interaction-Guided Exploration for Multi-Agent Monte Carlo Tree Search

Pages

Page
Executive summary
Claim 1: Low-dimensional nonlinear surrogate
Claim 2: Regret bound (Theorem 3.5)
Claim 3: Efficiency separation (Theorem 3.7)
Claim 4: MatGame 8-agent 10-action nonlinear
Claim 5: SMAC win rate
Claim 6: SMACv2 win rate + ablations
Conclusion