Research map / Learning theory and algorithms / Combinatorial and search optimization
Game Solving with Online Fine-Tuning
Ti-Rong Wu, Hung Guei, Ting Han Wei, Chung-Chin Shih, Jui-Te Chin, I-Chen Wu
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 online fine-tuning with critical positions | distributed PCN solver · 16 7x7 Killall-Go openings | total solving time (s): 1,021,692 -> 240,556 | Improved |
| add online fine-tuning with solved positions only | distributed PCN solver · 16 7x7 Killall-Go openings | total solving time (s): 1,021,692 -> 538,092 | Improved |
| combine solved and critical positions | online-cp · 16 7x7 Killall-Go openings | total solving time (s): 240,556 -> 255,359 | No clear effect |
| fine-tune on critical positions instead of solved positions | online fine-tuning · 16 openings | total solving time (s): 538,092 -> 240,556 | Improved |
| online fine-tuning on the easiest openings SA and SC | online-cp · SA, SC | solving time | Hurt |
| update PCN for both manager and workers | online-cp vs baseline · opening KE | solving time (s): 18,771 -> 3,336 | Improved |
| update PCN for workers only (ablation) | online-cp-w · opening KE | solving time (s): 18,771 -> 54,360 | Hurt |
| update PCN for manager only (ablation) | online-cp-m · opening KE | solving time (s): 18,771 -> 5,445 | Improved |
| remove AND-player job assignment (ablation) | online-cp · opening DA | solving time (s): 30,313 -> 53,275 | Hurt |
| remove top-k selection (ablation) | online-cp · opening SE | solving time (s): 12,465 -> 36,093 | Hurt |
| remove top-k selection (ablation) | online-cp · opening JC | solving time (s): 6,537 -> 6,520 | No clear effect |
| remove both AND-player assignment and top-k (ablation) | online-cp · four openings JC/KE/DA/SE | solving time: 6,537 -> 10,480 | Hurt |
Closest papers by meaning
- AlphaZero-based Proof Cost Network to Aid Game Solving (ICLR 2022 · similarity 0.86)
- Implicit Search via Discrete Diffusion: A Study on Chess (ICLR 2025 · similarity 0.83)
- Mastering Board Games by External and Internal Planning with Language Models (ICML 2025 · similarity 0.82)
- Mastering Zero-Shot Interactions in Cooperative and Competitive Simultaneous Games (ICML 2024 · similarity 0.82)
- On the Optimality of Dilated Entropy and Lower Bounds for Online Learning in Extensive-Form Games (NeurIPS 2024 · similarity 0.82)
- From Average-Iterate to Last-Iterate Convergence in Games: A Reduction and Its Applications (NeurIPS 2025 · similarity 0.82)
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.
