A Survey on Collaborating Small and Large Language Models for Performance, Cost-effectiveness, Cloud-edge Privacy, and Trustworthiness

TL;DR

This survey addresses the practical challenges of deploying large language models by examining how small language models can collaborate with LLMs across four goals: performance, cost-efficiency, cloud-edge privacy, and trustworthiness. It introduces a taxonomy and analyzes representative methods through two performance-oriented paradigms (guidance–generation and division–fusion), three cost-reduction stages (pre-training, tuning, and inference), and privacy/trust frameworks for cloud–edge setups. Key contributions include organizing existing work around collaboration objectives, detailing design paradigms like cascade routing, speculative decoding, and safety-guided decoding, and outlining open challenges such as interoperable ecosystems and formal privacy guarantees. The findings highlight practical pathways to build efficient, secure, and deployable SLM–LLM systems, with implications for real-world tasks ranging from domain-specific reasoning to privacy-preserving on-device inference and safety governance.

Abstract

Large language models (LLMs) have achieved remarkable progress across domains and applications but face challenges such as high fine-tuning costs, inference latency, limited edge deployability, and reliability concerns. Small language models (SLMs), with compact, efficient, and adaptable features, offer promising solutions. Building on this potential, recent research explores collaborative frameworks that integrate their complementary strengths, leveraging SLMs' specialization and efficiency with LLMs' generalization and reasoning to address diverse objectives across tasks and deployment scenarios. Motivated by these developments, this paper presents a systematic survey of SLM-LLM collaboration from the perspective of collaboration objectives. We propose a taxonomy covering four goals: performance enhancement, cost-effectiveness, cloud-edge privacy, and trustworthiness. Under this framework, we review representative methods, summarize design paradigms, and outline open challenges and future directions toward efficient and secure SLM-LLM collaboration. The collected papers are available at https://github.com/FairyFali/SLMs-Survey.

Figures (19)

• Figure 1: Detailed taxonomy of SLM–LLM collaboration, depicting the sub-taxonomy under the collaboration objectives of performance, cost-effectiveness, privacy, and trustworthiness.
• Figure 2: A Taxonomy of SLM-LLM Collaboration.
• Figure 3: Guidance-Generation Collaboration.
• Figure 4: Parallel Ensemble: Collective Intelligence.
• Figure 5: Sequential Cooperation: Division of Labor.