Paper Overview
Research area: Machine Learning
Authors: Qintong Xie, Edward Koh, Xavier Cadet
Published: 2025-06-11
arXiv: 2506.08262
Summary
Many real-world competitive systems require multiple decision-makers to act simultaneously under shared constraints, limited information, and repeated interactions—such as auctions, resource allocation, and security competitions. The authors study multi-turn simultaneous bidding as a controlled testbed for this class of problems and propose DNQ, a solver-in-the-loop equilibrium supervision framework for training bidding agents.
DNQ alternates between four stages:
1. Trajectory collection 2. Critic-based payoff estimation 3. Equilibrium computation 4. Policy imitation
A key contribution is a pairwise formulation that significantly reduces equilibrium-solving cost and training time. Experiments demonstrate that the pairwise approach scales to a larger number of agents.
Original Abstract
> We study multi-turn simultaneous bidding as a controlled testbed and propose DNQ, a solver-in-the-loop equilibrium supervision framework for training bidding agents. DNQ alternates between trajectory collection, critic-based payoff estimation, equilibrium computation, and policy imitation.
---
*Auto-collected on 2025-06-11.*