SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918450504400896 |
|---|---|
| author | Xie, Xin Xue, Dongyun Yao, Wuguannan Feng, Mingxiao Zhou, Wengang Qi, Xiang Li, Houqiang Zhang, Peng |
| author_facet | Xie, Xin Xue, Dongyun Yao, Wuguannan Feng, Mingxiao Zhou, Wengang Qi, Xiang Li, Houqiang Zhang, Peng |
| contents | LLM-powered systems require complex multi-step decision-making abilities to solve real-world tasks, yet current planning approaches face a trade-off between the high latency of inference-time search and the limited generalization of supervised fine-tuning. To address this limitation, we introduce \textbf{SGA-MCTS}, a framework that casts LLM planning as non-parametric retrieval. Offline, we leverage Monte Carlo Tree Search (MCTS) to explore the solution space and distill high-fidelity trajectories into State-Goal-Action (SGA) atoms. These atoms are de-lexicalized primitives that abstract concrete entities into symbolic slots, preserving reusable causal logic while discarding domain-specific noise. Online, a retrieval-augmented agent employs a hybrid symbolic-semantic mechanism to fetch relevant SGAs and re-ground them into the current context as soft reasoning hints. Empirical results on complex benchmarks demonstrate that this paradigm enables frozen, open-weights models to match the performance of SOTA systems (e.g., GPT-5) without task-specific fine-tuning. By effectively amortizing the heavy computational cost of search, SGA-MCTS achieves System 2 reasoning depth at System 1 inference speeds, rendering autonomous planning both scalable and real-time feasible. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_14712 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval Xie, Xin Xue, Dongyun Yao, Wuguannan Feng, Mingxiao Zhou, Wengang Qi, Xiang Li, Houqiang Zhang, Peng Artificial Intelligence LLM-powered systems require complex multi-step decision-making abilities to solve real-world tasks, yet current planning approaches face a trade-off between the high latency of inference-time search and the limited generalization of supervised fine-tuning. To address this limitation, we introduce \textbf{SGA-MCTS}, a framework that casts LLM planning as non-parametric retrieval. Offline, we leverage Monte Carlo Tree Search (MCTS) to explore the solution space and distill high-fidelity trajectories into State-Goal-Action (SGA) atoms. These atoms are de-lexicalized primitives that abstract concrete entities into symbolic slots, preserving reusable causal logic while discarding domain-specific noise. Online, a retrieval-augmented agent employs a hybrid symbolic-semantic mechanism to fetch relevant SGAs and re-ground them into the current context as soft reasoning hints. Empirical results on complex benchmarks demonstrate that this paradigm enables frozen, open-weights models to match the performance of SOTA systems (e.g., GPT-5) without task-specific fine-tuning. By effectively amortizing the heavy computational cost of search, SGA-MCTS achieves System 2 reasoning depth at System 1 inference speeds, rendering autonomous planning both scalable and real-time feasible. |
| title | SGA-MCTS: Decoupling Planning from Execution via Training-Free Atomic Experience Retrieval |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2604.14712 |