Research map / Probabilistic and scientific modeling / Variational and generative sampling
Variance Reduction in Monte Carlo Counterfactual Regret Minimization (VR-MCCFR) for Extensive Form Games Using Baselines
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 |
|---|---|---|---|
| Add state-action baselines and recursive bootstrapping | MCCFR on Leduc poker · Leduc poker | speedup over MCCFR: an order of magnitude | Helped |
| Add state-action baselines and recursive bootstrapping, and use CFR+ | MCCFR on Leduc poker · Leduc poker | iterations to reach the same exploitability: 250-times | Helped |
| Combine sampling with CFR+ using the proposed variance reduction | MCCFR on Leduc poker · Leduc poker | speedup over MCCFR: two orders of magnitude | Helped |
| Use state-action baselines instead of state baselines (ablation) | VR-MCCFR+ on Leduc poker · Leduc poker | convergence performance | Helped |
| Add recursive bootstrapping (ablation) | VR-MCCFR+ on Leduc poker · Leduc poker | convergence performance | Helped |
| Use both recursive bootstrapping and an action-dependent baseline (ablation) | VR-MCCFR+ on Leduc poker · Leduc poker | convergence performance | Helped |
| Use an oracle baseline with true counterfactual values instead of a learned baseline (ablation) | VR-MCCFR+ on Leduc poker · Leduc poker | exploitability | Helped |
| Learn the baseline online rather than use the oracle baseline (ablation) | VR-MCCFR+ on Leduc poker · Leduc poker | exploitability | No clear effect |
| Use CFR+ rather than standard regret matching in the sampling algorithm (ablation) | MCCFR on Leduc poker · Leduc poker | convergence performance: For the first $10^6$ iterations; approximately $10^7$ iterations | Mixed |
Closest papers by meaning
- Stochastic Regret Minimization in Extensive-Form Games (ICML 2020 · similarity 0.90)
- Accelerating Nash Equilibrium Convergence in Monte Carlo Settings Through Counterfactual Value Based Fictitious Play (NeurIPS 2024 · similarity 0.90)
- Dynamic Discounted Counterfactual Regret Minimization (ICLR 2024 · similarity 0.89)
- GAE Falls Short in Imperfect-Information Self-Play Reinforcement Learning (NeurIPS 2026 · similarity 0.86)
- Variance-Optimal State Resampling for Reinforcement Learning with Verifiable Rewards (NeurIPS 2026 · similarity 0.86)
- Exploring Starts Are Not Enough: Counterexamples and a Fix for Monte Carlo Exploring Starts (NeurIPS 2026 · similarity 0.86)
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.