monte carlo tree search
Imagine you're trying to decide the best move in a game by simulating many possible futures, but you focus most of your effort on the most promising paths instead of exploring every random branch equally. Monte Carlo tree search does exactly that: it builds a tree of decisions, simulates random playouts from each node, and uses those results to guide which paths to explore more deeply, balancing trying new moves with exploiting known good ones.
These papers all use MCTS as a core mechanism to solve problems where intelligent search over a large space is key, but each extends it in a different way. HyperMCTS reuses statistics across similar paths for planning, the AlphaZero chess system uses it for self-play training without human data, and the speculative decoding work uses it to find optimal compilation plans. A common theme is using MCTS to efficiently allocate computational resources, whether to improve planning accuracy, achieve strong chess performance with limited hardware, or accelerate GPU inference.
- Papers
- 5
- Released code
- 0
- First seen
- Sept 2026
- Latest
- Nov 2026
5 papers in the last two months, against 0 in the two before.
Start here
Three papers to read first, in the order they came out- 1Jointly co-compiling draft and target models boosts speculative decoding speed
Conjugate Compilation jointly optimizes draft and target models for speculative decoding, achieving up to 1.59x speedup on single GPUs.
Oct 2026 - 2Training shifts a small theorem prover toward fewer, larger proof steps
Reap trains a 1.7B-parameter prover that solves 79.9% of miniF2F-test via Monte Carlo search.
Nov 2026 - 3AI agents improve when their surrounding software is automatically redesigned
HarnessFlow achieves the highest mean test scores on six benchmarks by using MCTS-guided search to optimize LLM agent harnesses.
Nov 2026
How fast it is moving
Often studied with
reinforcement learningA machine learning method where an agent learns to make sequences of decisions by receiving rewards or penalties for its actions.2 togetherself-evolving agentAn agent that improves by reusing its own experience, memory, or self-generated feedback, with success depending on verification and design.2 togethercredit assignmentDetermining which actions or steps in a sequence contributed to a final outcome, often for reinforcement learning.1 togetherThe papers
Most central to this idea first, not most recent.
- Big Techcs.AI
AI agents improve when their surrounding software is automatically redesigned
HarnessFlow achieves the highest mean test scores on six benchmarks by using MCTS-guided search to optimize LLM agent harnesses.
South China University of Technology, Fudan University · Nov 2026
- Top Universitycs.AI
Training shifts a small theorem prover toward fewer, larger proof steps
Reap trains a 1.7B-parameter prover that solves 79.9% of miniF2F-test via Monte Carlo search.
Peking University, IQuest Research · Nov 2026
- Big Techcs.LG
Jointly co-compiling draft and target models boosts speculative decoding speed
Conjugate Compilation jointly optimizes draft and target models for speculative decoding, achieving up to 1.59x speedup on single GPUs.
University of California, San Diego, Massachusetts Institute of Technology · Oct 2026
- Big Techcs.AI
Hypergraphs help language agents reuse feedback across planning paths
HyperMCTS uses a hypergraph to share decision outcomes across trajectories, improving MCTS efficiency and accuracy for LLM agents.
University of Florida, Amazon · Sept 2026
- Independentcs.LG
A compact AlphaZero-style chess system reaches 3,251 Elo after 2.5 days on a single GPU node
An engineered AlphaZero-style chess system achieves 3251 Elo in 2.5 days on one 8-GPU node.
Oct 2026
Concepts are extracted from each paper and reused across the corpus, so this page grows on its own as the desk reads.