Research map / Learning theory and algorithms / Submodular and game algorithms
General search techniques without common knowledge for imperfect-information games, and application to superhuman Fog of War chess
Brian Hu Zhang, Tuomas Sandholm
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 purification of the final mixed strategy (ablation) | Obscuro · Obscuro self-play, 5 s/move | match score (%): 70.2% for full Obscuro vs version without | Improved |
| add KLUSS (unfreezing strategies) instead of 1-KLSS (ablation) | Obscuro · Obscuro self-play, 5 s/move | match score (%): 58.0% for full Obscuro vs version without | Improved |
| use one-sided GT-CFR expansion instead of two-sided (ablation) | Obscuro · Obscuro self-play, 5 s/move | match score (%): 53.3% for full Obscuro vs version without | Improved |
| use non-uniform Resolve root distribution instead of uniform (ablation) | Obscuro · Obscuro self-play, 5 s/move | match score (%): 53.3% for full Obscuro vs version without | Improved |
| replace Stockfish 14 evaluation with simple hand-crafted evaluation (ablation) | Obscuro · Obscuro vs SE Obscuro | match score (%): 81.9% for Obscuro with Stockfish eval | Hurt |
| GT-CFR search with simple evaluation vs prior ZS21 AI | SE Obscuro · SE Obscuro vs ZS21 | match score (%): 55.0% for SE Obscuro | Improved |
| double search time per move (e.g. 1/16 s to 1/8 s) (ablation) | Obscuro · Obscuro self-play | match score (%): 56.4% at 1/8 s vs 1/16 s; 52.3% at 16 s vs 8 s | Improved |
Closest papers by meaning
- Human-Aligned Chess With a Bit of Search (ICLR 2025 · similarity 0.86)
- The Update-Equivalence Framework for Decision-Time Planning (ICLR 2024 · similarity 0.85)
- Opponent-Model Search in Games with Incomplete Information (AAAI 2024 · similarity 0.85)
- Designing Skill-Compatible AI: Methodologies and Frameworks in Chess (ICLR 2024 · similarity 0.85)
- Mastering Board Games by External and Internal Planning with Language Models (ICML 2025 · similarity 0.84)
- Amortized Planning with Large-Scale Transformers: A Case Study on Chess (NeurIPS 2024 · similarity 0.83)
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.
