Efficient LLM Collaboration via Planning

[Submitted on 13 Jun 2025 (v1), last revised 25 Aug 2026 (this version, v5)]

View PDF HTML (experimental)

Abstract:Recently, large language models (LLMs) have demonstrated strong performance, ranging from simple to complex tasks. However, while large models achieve remarkable results across diverse tasks, they often incur substantial monetary inference cost, making frequent use impractical for many applications. In contrast, small models are often freely available and easy to deploy locally, but their performance on complex tasks remains limited. This trade-off raises a natural question: how can small and large models efficiently collaborate to combine their complementary strengths? To bridge this trade-off, we propose COPE, a test-time collaboration framework. A planner model first generates a plan that serves as a lightweight intermediate that guides a downstream executor model. Small and large models take turns acting as planner and executor, exchanging plans in a multi-stage cascade to collaboratively solve tasks. Through comprehensive experiments on benchmarks spanning mathematical reasoning, code generation, open-ended tasks, and agent tasks, we demonstrate that COPE achieves performance comparable to large proprietary models, while drastically reducing the inference API cost. These results highlight planning as an effective prior for cost-efficient inference.

Submission history

From: Jonghoon Lee [view email]
[v1] Fri, 13 Jun 2025 08:35:50 UTC (181 KB)
[v2] Sat, 27 Sep 2025 14:19:07 UTC (281 KB)
[v3] Fri, 16 Jan 2026 15:28:18 UTC (288 KB)
[v4] Mon, 11 May 2026 07:58:40 UTC (272 KB)
[v5] Tue, 25 Aug 2026 02:33:49 UTC (1,163 KB)