Efficient Agentic Reasoning Through Self-Regulated Simulative Planning

arXiv 2026 49 method

TLDR

Proposes self-regulated simulative planning with three systems, using LLM as world model, achieving efficient reasoning competitive with much larger models.

Reasoning

Strengths: novel three-system decomposition for efficient planning, self-regulation reduces token use, competitive results with smaller models. Weaknesses: reliance on LLM as world model may limit generalization; abstract lacks detailed methodology and limitations.

Read-first score

Read-first score 49, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 30.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Citation impact 18%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.83040786

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=analysis,evaluation

Topical relevance 29%

42.9

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FoundationFrontierBridge

Rank sensitivity

Stability: volatile; rank range: 337.

Keyword Scores

world model

world dynamics prediction

world simulator

model-based reinforcement learning world model

generative world model

interactive world model

video world model

Deep Analysis

Innovations

Decomposition of decision-making into three systems: simulative reasoning (System II), self-regulation (System III), and reactive execution (System I)
SR^2^AM (Self-Regulated Simulative Reasoning Agentic LLM) realizing these as distinct stages within an LLM's chain-of-thought, with the LLM as world model
Two instantiations: v0.1 (prompted multi-module) and v1.0 (reconstructed from traces of pretrained reasoning LLMs, trained via supervised then RL)
Self-regulation decides when and how deeply to plan, leading to efficient token use

Methodology

The paper develops SR^2^AM, an agentic LLM that decomposes reasoning into three systems: reactive execution (System I), simulative reasoning (System II) using a world model, and self-regulation (System III) via a learned configurator. Two instantiations are explored: v0.1 uses a prompted multi-module system, while v1.0 reconstructs structured plans from traces of pretrained reasoning LLMs and is trained via supervised learning followed by reinforcement learning. Evaluation is conducted across math, science, tabular analysis, and web information seeking tasks.

Key Results

v0.1-8B and v1.0-30B achieve Pass@1 competitive with 120-355B and 685B-1T parameter systems respectively, while v1.0-30B uses 25.8-95.3% fewer reasoning tokens than comparable agentic LLMs. RL increases average planning horizon by 22.8% while planning frequency grows only 2.0%, showing it learns to plan further ahead rather than more often.

Resources

Paper