Research map / Reinforcement learning / Visual and representation RL
Iterative Empirical Game Solving via Single Policy Best Response
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use Mixed-Oracles with Q-Mixing to reuse responses to previously encountered opponent policies and train only against the newest opponent | PSRO · Gathering-Small | SumRegret: approximately 30% less regret | Helped |
| Use Mixed-Oracles to train against the newest opponent policy and transfer prior responses | PSRO · Gathering-Small | training simulations: a quarter of the number of simulations | Helped |
| Use Mixed-Oracles to train against the newest opponent policy and transfer prior responses | PSRO · Leduc Poker | SumRegret | No clear effect |
| Use Mixed-Oracles to train against the newest opponent policy and transfer prior responses | PSRO · Leduc Poker | training simulations: dramatically fewer cumulative timesteps | Helped |
| Use Mixed-Opponents with Q-Mixing to construct a single representative opponent policy | PSRO · Gathering-Open | SumRegret: after 25 epochs; 50% less regret | Helped |
| Use Mixed-Opponents with Q-Mixing to construct a single representative opponent policy | PSRO · Gathering-Open | SumRegret | Helped |
| Use Mixed-Opponents to train against a single representative opponent policy | PSRO · Leduc Poker | SumRegret | No clear effect |
| Use Mixed-Opponents to train against a single representative opponent policy | PSRO · Leduc Poker | training simulations: dramatically fewer timesteps | Helped |
| Use Mixed-Opponents, which aggregates the joint opponent mixture into one policy | PSRO · Gathering-Open, 3 players | training time: half of the time | Helped |
| Use Mixed-Opponents with shared RL hyperparameters | PSRO · Gathering-Small, pure-hparams | SumRegret: half the simulation timesteps | Helped |
| Use Mixed-Oracles and Mixed-Opponents with shared RL hyperparameters | PSRO · Gathering-Small | SumRegret: in fewer timesteps | Helped |
Closest papers by meaning
- Explicit Exploration for High-Welfare Equilibria in Game-Theoretic Multiagent Reinforcement Learning (ICML 2025 · similarity 0.91)
- Global Policy-Space Response Oracles for Two-Player Zero-Sum Games (ICML 2026 · similarity 0.91)
- Toward Optimal Policy Population Growth in Two-Player Zero-Sum Games (ICLR 2024 · similarity 0.90)
- A-PSRO: A Unified Strategy Learning Method with Advantage Metric for Normal-form Games (ICML 2025 · similarity 0.88)
- Oracle-Efficient Reinforcement Learning for Max Value Ensembles (NeurIPS 2024 · similarity 0.88)
- Prompted Policy Search: Reinforcement Learning through Linguistic and Numerical Reasoning in LLMs (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.