# ArborKV: Structure-Aware KV Cache Management for Scaling Tree-based LLM Reasoning

Source: [ArborKV: Structure-Aware KV Cache Management for Scaling Tree-based LLM Reasoning](<https://www.sota2.com/research/paper/6a0fe32cf984739ed6739f06>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-23T05:00:04.712Z

Authors: Yeqiu Chen, Ziyan Liu, Zhenxin Huang, Runquan Gui, Hong Wang, Lei Liu

Published: 2026-05-21T07:40:57.000Z

## Abstract

Recent progress in LLM reasoning has increasingly shifted from single-pass generation to explicit search over intermediate reasoning states. Tree-of-Thoughts (ToT) organizes inference to tree-structured search with branching and backtracking, but it substantially amplifies the Key--Value (KV) cache: retaining KV states for a frontier of partial trajectories quickly becomes a memory bottleneck that limits throughput and constrains search depth and width under fixed hardware budgets. We address this challenge by observing that KV reuse in ToT-style inference is governed by search dynamics: near-term decoding depends primarily on the active branch and its ancestors, whereas inactive subtrees have low short-term reuse probability yet must remain recoverable for backtracking. Motivated by this, we propose ArborKV, a structure-aware eviction framework that couples a lightweight value estimator with a tree-aware allocation policy, and performs purely token-extractive eviction with lazy rehydration to support revisits. Experiments on ToT-style reasoning benchmarks show that ArborKV achieves up to ~4x peak KV-memory reduction while preserving near-full-retention accuracy, enabling larger search configurations under fixed device budgets that would otherwise run out of memory.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Mathematical Reasoning on AIME (Accuracy, Peak Memory)](<https://www.sota2.com/research/sota/mathematical-reasoning-on-aime-accuracy-peak-memory.md>): Mathematical Reasoning; AIME; Accuracy: 33.8.

- [Mathematical Reasoning on MATH 500 (Accuracy, Memory Usage)](<https://www.sota2.com/research/sota/mathematical-reasoning-on-math-500-accuracy-memory-usage.md>): Mathematical Reasoning; MATH 500; Accuracy: 77.6.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.22106>)

[PDF](<https://arxiv.org/pdf/2605.22106.pdf>)