Research map / Learning theory and algorithms / Combinatorial and search optimization
Learning to search with MCTSnets
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 learned vector-valued node statistics, simulation policy, and backup instead of standard UCT search | MCTS on Sokoban · Sokoban | success ratio: ≈ 30% -> 84±1 % | Helped |
| Use learned search rather than a PUCT-like rule with pre-learned policy and value networks | MCTS on Sokoban · Sokoban | success ratio: ≈ 65% -> 84±1 % | Helped |
| Add a gated residual connection to the backup network instead of using a simple MLP (ablation) | MCTSnet · Sokoban | success ratio | Helped |
| Use a non-residual backup network instead of a gated residual backup (ablation) | MCTSnet · Sokoban, M>25 simulations | numerical stability | Hurt |
| Learn a simple simulation policy instead of using a uniform random simulation policy (ablation) | MCTSnet · Sokoban | success ratio | Helped |
| Use the proposed credit-assignment scheme with a modulated simulation policy (ablation) | MCTSnet · Sokoban | success ratio | Helped |
| Set credit-assignment discount γ below 1 rather than optimizing the true terminal loss with γ=1 (ablation) | MCTSnet · Sokoban | success ratio | Helped |
| Use only the policy-prior term (set w1=0) instead of the best learned modulated policy (ablation) | MCTSnet · Sokoban | success ratio | Hurt |
| Increase the number of simulations (ablation) | MCTSnet · Sokoban | success ratio | Helped |
| Replace the real environment model with the sham model T(s,a)=s and r(s,a)=0 (ablation) | MCTSnet · Sokoban | success ratio | Hurt |
Closest papers by meaning
- Monte Carlo Tree Search in the Presence of Transition Uncertainty (AAAI 2024 · similarity 0.87)
- Monte Carlo Tree Diffusion for System 2 Planning (ICML 2025 · similarity 0.86)
- Position: Rethinking Post-Hoc Search-Based Neural Approaches for Solving Large-Scale Traveling Salesman Problems (ICML 2024 · similarity 0.86)
- Compositional Monte Carlo Tree Diffusion for Extendable Planning (NeurIPS 2025 · similarity 0.86)
- ExACT: Teaching AI Agents to Explore with Reflective-MCTS and Exploratory Learning (ICLR 2025 · similarity 0.86)
- Threshold UCT: Cost-Constrained Monte Carlo Tree Search with Pareto Curves (AAAI 2025 · 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.