← All notes

Short answer

Language Agent Tree Search (LATS) has the agent explore several action branches with Monte Carlo Tree Search, scoring them with model-based values and self-reflection plus environment feedback. It reached 92.7% pass@1 on HumanEval. A survey of agent architectures names planning, execution and reflection as key phases for robust agents. [S1][S2]

Claims

• id: C1 claim: LATS integrates Monte Carlo Tree Search into language agents. source: S1 evidence: "we integrate Monte Carlo Tree Search into LATS to enable LMs as agents"
• id: C2 claim: LATS achieved 92.7% pass@1 on HumanEval. source: S1 evidence: "LATS achieves state-of-the-art pass@1 accuracy (92.7%) for programming on HumanEval"
• id: C3 claim: Planning, execution and reflection are key phases for robust agents. source: S2 evidence: "key phases for planning, execution, and reflection that enable robust AI agent systems"

Sources

• [S1] [2310.04406] Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models. arXiv. https://arxiv.org/abs/2310.04406 . License: CC BY 4.0. Accessed 2026-11-04.
• [S2] [2404.11584] The Landscape of Emerging AI Agent Architectures for Reasoning, Planning, and Tool Calling: A Survey. arXiv. https://arxiv.org/abs/2404.11584 . License: CC BY 4.0. Accessed 2026-11-04.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted).

@misc{iil_note_cqqked57,
  author = {{iinventedlight}},
  title = {{Tree search for agent planning (LATS)}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-04},
  url = {https://iinventedlight.com/library/notes/ao-planning-tree-search/},
  note = {Permanent ID iil:note:cqqked57. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt. This page in Markdown: /library/notes/ao-planning-tree-search.md.