Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cached Model-as-a-Resource: Provisioning Large Language Model Agents for Edge Intelligence in Space-air-ground Integrated Networks

Minrui Xu, Dusit Niyato, Hongliang Zhang, Jiawen Kang, Zehui Xiong, Shiwen Mao, Zhu Han

TL;DR

This work tackles provisioning LLM agent services across space-air-ground integrated networks by treating cached LLM models as a resource and introducing an AoT-based, least AoT caching strategy to manage context-window efficiency at edge servers. It pairs this with a DRL-guided, strategy-proof market mechanism (DQMSB) that uses a price-scaling factor to maximize total surplus while avoiding adverse selection. The approach yields substantial gains in efficiency (notably around 23% higher surplus over baselines) and guarantees anonymity, strategy-proofness, and adverse-selection-freeness. The combination of AoT-aware caching and DRL-based auction design offers a practical framework for scalable, low-latency edge intelligence in SAGINs with satellites relaying requests to cloud datacenters.

Abstract

Edge intelligence in space-air-ground integrated networks (SAGINs) can enable worldwide network coverage beyond geographical limitations for users to access ubiquitous and low-latency intelligence services. Facing global coverage and complex environments in SAGINs, edge intelligence can provision approximate large language models (LLMs) agents for users via edge servers at ground base stations (BSs) or cloud data centers relayed by satellites. As LLMs with billions of parameters are pre-trained on vast datasets, LLM agents have few-shot learning capabilities, e.g., chain-of-thought (CoT) prompting for complex tasks, which raises a new trade-off between resource consumption and performance in SAGINs. In this paper, we propose a joint caching and inference framework for edge intelligence to provision sustainable and ubiquitous LLM agents in SAGINs. We introduce "cached model-as-a-resource" for offering LLMs with limited context windows and propose a novel optimization framework, i.e., joint model caching and inference, to utilize cached model resources for provisioning LLM agent services along with communication, computing, and storage resources. We design "age of thought" (AoT) considering the CoT prompting of LLMs, and propose a least AoT cached model replacement algorithm for optimizing the provisioning cost. We propose a deep Q-network-based modified second-bid (DQMSB) auction to incentivize network operators, which can enhance allocation efficiency by 23% while guaranteeing strategy-proofness and free from adverse selection.

Cached Model-as-a-Resource: Provisioning Large Language Model Agents for Edge Intelligence in Space-air-ground Integrated Networks

TL;DR

This work tackles provisioning LLM agent services across space-air-ground integrated networks by treating cached LLM models as a resource and introducing an AoT-based, least AoT caching strategy to manage context-window efficiency at edge servers. It pairs this with a DRL-guided, strategy-proof market mechanism (DQMSB) that uses a price-scaling factor to maximize total surplus while avoiding adverse selection. The approach yields substantial gains in efficiency (notably around 23% higher surplus over baselines) and guarantees anonymity, strategy-proofness, and adverse-selection-freeness. The combination of AoT-aware caching and DRL-based auction design offers a practical framework for scalable, low-latency edge intelligence in SAGINs with satellites relaying requests to cloud datacenters.

Abstract

Edge intelligence in space-air-ground integrated networks (SAGINs) can enable worldwide network coverage beyond geographical limitations for users to access ubiquitous and low-latency intelligence services. Facing global coverage and complex environments in SAGINs, edge intelligence can provision approximate large language models (LLMs) agents for users via edge servers at ground base stations (BSs) or cloud data centers relayed by satellites. As LLMs with billions of parameters are pre-trained on vast datasets, LLM agents have few-shot learning capabilities, e.g., chain-of-thought (CoT) prompting for complex tasks, which raises a new trade-off between resource consumption and performance in SAGINs. In this paper, we propose a joint caching and inference framework for edge intelligence to provision sustainable and ubiquitous LLM agents in SAGINs. We introduce "cached model-as-a-resource" for offering LLMs with limited context windows and propose a novel optimization framework, i.e., joint model caching and inference, to utilize cached model resources for provisioning LLM agent services along with communication, computing, and storage resources. We design "age of thought" (AoT) considering the CoT prompting of LLMs, and propose a least AoT cached model replacement algorithm for optimizing the provisioning cost. We propose a deep Q-network-based modified second-bid (DQMSB) auction to incentivize network operators, which can enhance allocation efficiency by 23% while guaranteeing strategy-proofness and free from adverse selection.
Paper Structure (27 sections, 3 theorems, 30 equations, 6 figures, 2 algorithms)

This paper contains 27 sections, 3 theorems, 30 equations, 6 figures, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Edge Intelligence in Space-air-ground Integrated Networks
  4. Large Language Models for Edge Intelligence
  5. Auction Design for SAGINs
  6. System Model
  7. Coverage Time Model
  8. Communication Model
  9. Model Caching for LLM Agent Services
  10. Chain-of-Thought Inference Model
  11. Problem Formulation, Caching Algorithm Design, and Market Design
  12. Cost Structure
  13. Edge Inference Cost
  14. Cloud Inference Cost
  15. Problem Formulation of Ground Base Stations
  16. ...and 12 more sections

Key Result

Theorem 1

Considering a collection of $c_i$ varying length CoT examples, which are generated from the intention $\theta^\star$ with the optimal context $c^\star$ sampled from $q_m(c)$ that satisfies Assumption context. Furthermore, let $d_{i,0}$ be the input message or task sampled from $q(\cdot | \theta^\sta where $\eta = 2 \frac{\epsilon(d_{i,0})}{1-\epsilon(d_{i,0})}$ depends on the ambiguity of the inpu

Figures (6)

  • Figure 1: Joint caching and inference framework for provisioning large language model (LLM) agents in SAGINs.
  • Figure 2: The workflow of the joint caching and inference framework for provisioning LLM agents with cached models.
  • Figure 3: Performance of model caching algorithms under different system settings.
  • Figure 4: Model performance under different vanishing factors.
  • Figure 5: The convergence of the proposed DQMSB auction.
  • ...and 1 more figures

Theorems & Definitions (8)

  • Definition 1: $\epsilon$-ambiguity jiang2023latent
  • Definition 2
  • Theorem 1
  • proof
  • Lemma 1
  • proof
  • Theorem 2
  • proof