The Year in AI Papers
The year/Independent research

Paper 2601.09259

MAXS: Meta-Adaptive Exploration with LLM Agents

Jian Zhang, Zhiyuan Wang, Zhangqi Wang, Yu He, Haoran Luo, li yuan, Lingling Zhang, Rui Mao, Qika Lin, Jun Liu

Published
Jan 2026
Research lab
Independent
Citations
1
GitHub
Not linked

01 In brief

Summary

The paper introduces MAXS, a meta-adaptive reasoning framework for LLM agents that addresses two issues: locally myopic generation and trajectory instability.

MAXS uses a lookahead strategy to extend reasoning paths a few steps ahead, estimating the advantage value of tool usage, and combines step consistency variance and inter-step trend slopes to select stable, consistent, and high-value reasoning steps.

A trajectory convergence mechanism halts rollouts once path consistency is achieved, balancing resource efficiency and global effectiveness.

Experiments across three backbones (MiMo-VL-7B, Qwen2.5-VL-7B, Qwen2.5-VL-32B) and five datasets (MathVista, OlympiadBench, EMMA, TheoremQA, MATH) show MAXS consistently outperforms baselines (CoT, ToT, MCTS, Guided Decoding, ϕ-Decoding) in both performance and inference efficiency.

For example, on MiMo-VL-7B, MAXS achieves 63.46% average accuracy, 6.42% higher than ToT, while using significantly fewer tokens (9.86×10^8 vs.

ToT's 6.40×10^10).

Ablations confirm the lookahead module is essential, the advantage score dominates value estimation, and trajectory convergence improves efficiency with minimal accuracy loss.

The 4-step lookahead offers the best accuracy-cost trade-off, and both code and search tools are complementary, with code being critical for symbolic reasoning.

02 From the paper

Abstract

Large Language Model (LLM) Agents exhibit inherent reasoning abilities through the collaboration of multiple tools. However, during agent inference, existing methods often suffer from (i) locally myopic generation, due to the absence of lookahead, and (ii) trajectory instability, where minor early errors can escalate into divergent reasoning paths. These issues make it difficult to balance global effectiveness and computational efficiency. To address these two issues, we propose meta-adaptive exploration with LLM agents https://github.com/exoskeletonzj/MAXS, a meta-adaptive reasoning framework based on LLM Agents that flexibly integrates tool execution and reasoning planning. MAXS employs a lookahead strategy to extend reasoning paths a few steps ahead, estimating the advantage value of tool usage, and combines step consistency variance and inter-step trend slopes to jointly select stable, consistent, and high-value reasoning steps. Additionally, we introduce a trajectory convergence mechanism that controls computational cost by halting further rollouts once path consistency is achieved, enabling a balance between resource efficiency and global effectiveness in multi-tool reasoning. We conduct extensive empirical studies across three base models (MiMo-VL-7B, Qwen2.5-VL-7B, Qwen2.5-VL-32B) and five datasets, demonstrating that MAXS consistently outperforms existing methods in both performance and inference efficiency. Further analysis confirms the effectiveness of our lookahead strategy and tool usage.