Hive: A Multi-Agent Infrastructure for Algorithm- and Task-Level Scaling

2026-04-19

AI Summary

This work addresses the inefficiencies in existing large language model (LLM) agent systems, which suffer from redundant reasoning and suboptimal resource scheduling when scaling at both the algorithmic and task levels. To overcome these limitations, the authors propose Hive, a multi-agent infrastructure that enables joint algorithmic and task-level scaling for the first time. Hive employs a declarative frontend to specify agent behaviors and introduces a Logits Cache mechanism in the backend to eliminate redundant sampling. Furthermore, it incorporates an agent-aware scheduler to optimize resource allocation. Experimental results demonstrate that Logits Cache achieves a 1.11–1.76× speedup in resampling, while the agent-aware scheduler reduces hotspot miss rates by 33%–51%, substantially enhancing scalability and parallel efficiency for complex tasks.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Multiagent Systems: Agent Communication

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large language models are increasingly deployed as complex agentic systems that scale with task complexity. While prior work has extensively explored model- and system-level scaling, algorithm- and task-level scaling remain largely unaddressed, constraining the full potential of agentic systems. At the algorithm level, allocating additional inference-time computation can enhance workflow capacity but introduces cross-path redundancy: overlapping computations across multiple reasoning branches. At the task level, complex tasks can be decomposed into subproblems and delegated across multiple agents for improved scalability and parallelism. However, existing infrastructures' scheduling is unaware of the existence of multiple agents, missing opportunities to optimize resource allocation. We propose Hive, a multi-agent infrastructure that enables algorithm- and task-level scaling. Hive features a description frontend that captures per-agent behavior and supports test-time scaling algorithms. Leveraging this specification, our backend introduces two key mechanisms: Logits Cache that reuses intermediate logits across redundant sampling paths to mitigate cross-path redundancy at the algorithm level, and Agent-Aware Scheduling that efficiently allocates compute and KV-cache resources according to agent contributions at the task level. Experiments show that Logits Cache achieves an average speedup of $1.11\times$-$1.76\times$ for re-sampling, and Agent-Aware Scheduling reduces the hotspot miss rate by $33\%$-$51\%$.

Problem

• algorithm-level scaling
• task-level scaling
• cross-path redundancy
• multi-agent systems
• resource allocation

Innovation

• multi-agent infrastructure
• algorithm-level scaling
• task-level scaling
• logits caching
• agent-aware scheduling