Lemon Agent Technical Report

2026-02-06

AI Summary

This work addresses the inherent limitations of large language model agents in resource efficiency, context management, and multimodal perception by proposing a multi-agent collaborative system built upon the AgentCortex framework. The system features a two-tier adaptive scheduling architecture that dynamically optimizes computational resource allocation, coupled with a three-level progressive context compression strategy and a self-evolving memory mechanism to enable efficient coordination between global planning and local execution. Furthermore, it integrates an enhanced Model Context Protocol (MCP) toolset to substantially improve information utilization and execution performance on complex, long-horizon tasks. Experimental results demonstrate that the system achieves a 91.36% accuracy on the GAIA benchmark and ranks first on the xBench-DeepSearch leaderboard with a score exceeding 77.

Technology Category

• Multiagent Systems: Agent Communication
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• 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

Abstract

Recent advanced LLM-powered agent systems have exhibited their remarkable capabilities in tackling complex, long-horizon tasks. Nevertheless, they still suffer from inherent limitations in resource efficiency, context management, and multimodal perception. Based on these observations, Lemon Agent is introduced, a multi-agent orchestrator-worker system built on a newly proposed AgentCortex framework, which formalizes the classic Planner-Executor-Memory paradigm through an adaptive task execution mechanism. Our system integrates a hierarchical self-adaptive scheduling mechanism that operates at both the overall orchestrator layer and workers layer. This mechanism can dynamically adjust computational intensity based on task complexity. It enables orchestrator to allocate one or more workers for parallel subtask execution, while workers can further improve operational efficiency by invoking tools concurrently. By virtue of this two-tier architecture, the system achieves synergistic balance between global task coordination and local task execution, thereby optimizing resource utilization and task processing efficiency in complex scenarios. To reduce context redundancy and increase information density during parallel steps, we adopt a three-tier progressive context management strategy. To make fuller use of historical information, we propose a self-evolving memory system, which can extract multi-dimensional valid information from all historical experiences to assist in completing similar tasks. Furthermore, we provide an enhanced MCP toolset. Empirical evaluations on authoritative benchmarks demonstrate that our Lemon Agent can achieve a state-of-the-art 91.36% overall accuracy on GAIA and secures the top position on the xbench-DeepSearch leaderboard with a score of 77+.

Problem

• resource efficiency
• context management
• multimodal perception
• long-horizon tasks
• LLM-powered agents

Innovation

• multi-agent system
• adaptive scheduling
• progressive context management
• self-evolving memory
• AgentCortex