Evaluating Temporal Semantic Caching and Workflow Optimization in Agentic Plan-Execute Pipelines

arXiv:2605.20630 · 2026-05-21 공개 · arXiv · PDF

latency-optimization assetopsbench workflow-optimization llm-caching plan-execute-pipeline industrial-agents cache-hit-speedup parameter-rich-queries

Abstract

Industrial asset operations workflows are latency-sensitive because a single user query may require coordination over sensor data, work orders, failure modes, forecasting tools, and domain-specific agents. We evaluate this problem on AssetOpsBench (AOB), an industrial agent benchmark whose plan-execute pipeline exposes repeated overhead from tool discovery, LLM planning, MCP tool execution, and final summarization. Existing LLM caching techniques such as KV-cache reuse and embedding-based semantic caching were designed for chatbot serving and break down when output validity depends on time, asset, or sensor parameters. We propose two complementary optimization layers for AOB plan-execute pipelines: a temporal semantic cache and a set of MCP workflow optimizations combining disk-backed tool-discovery caching and dependency-aware parallel step execution. MCP workflow optimizations corresponded to a 1.67x speedup and reduced median end-to-end latency by about 40.0% while the temporal-cache benchmark achieved a median of 30.6x speedup on cache hits. Beyond the speedup, our results expose a concrete failure mode of pure semantic caching for parameter-rich industrial queries, providing a critical analysis of how caching choices interact with evaluation correctness in MCP-backed agent benchmarks.

한국어 요약

📋 한 줄 요약

**[Agentic Pipeline / Temporal Caching]** AssetOpsBench에서 temporal semantic cache + MCP workflow 최적화 평가 — MCP workflow 최적화로 1.67× 가속·median end-to-end latency 약 40% 감소, temporal cache hit 시 median 30.6× 가속, 순수 semantic caching의 parameter-rich 산업 쿼리 실패 모드 노출.

🎯 핵심 기여도

💡 핵심 아이디어

산업 agentic pipeline의 latency는 단순 chatbot용 캐싱으로는 줄일 수 없고 — 시간·자산·센서 파라미터에 따라 output validity가 바뀌기 때문 — temporal-aware semantic cache와 dependency-aware parallel MCP workflow를 결합해야만 의미 있는 가속이 가능하며, parameter-rich 쿼리의 평가 correctness와 caching 선택의 상호작용을 분석할 필요가 있다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: 산업 agentic 시스템의 캐싱 설계 가이드 제시, 일반 LLM 캐싱과 산업 도메인의 구조적 차이 정량 노출, MCP workflow 최적화의 일반 패턴(parallel execution + tool discovery cache) 제공. **한계**: AOB 단일 벤치마크 검증, temporal cache의 staleness vs hit rate trade-off 가이드 추가 분석 여지, MCP 외 다른 agent 프레임워크 일반화는 별도 검증.

🚀 실용적 활용