AI-NativeBench: Raw Experimental Data

This repository contains the raw experimental data for the paper "AI-NativeBench: An Open-Source White-Box Agentic Benchmark Suite for AI-Native Systems".

The dataset provides a comprehensive "white-box" view of AI-Native system execution, capturing not just final outcomes but also the internal execution dynamics via distributed tracing. It covers 7 Large Language Models (LLMs) evaluated across 8 realistic applications and 21 system architectural variants.

📂 Dataset Overview

The data is organized hierarchically by Model $\rightarrow$ Application & Architecture.

1. Supported Models

The root directory contains folders for the 7 LLMs evaluated in the study:

GPT-5
GPT-4o-mini
DeepSeek-V3.1
DeepSeek-R1
Gemini-2.5-flash
Gemini-2.5-flash-nothinking
Qwen3-235b

2. Directory Structure

Inside each model folder, you will find the following subdirectories representing specific Task-Architecture combinations:

├── [Model_Name]/
│   ├── BookWriter-A2A                  # Content Generation (A2A Protocol)
│   ├── BookWriter-H_A2A                # Content Generation (Heterogeneous A2A)
│   ├── BookWriter-MCP                  # Content Generation (MCP Tools)
│   ├── EmailResponder                  # Communication (Pure Framework)
│   ├── EmailResponder-MCP              # Communication (MCP Tools)
│   ├── GameBuilder                     # Software Engineering (Pure Framework)
│   ├── GameBuilder-MCP                 # Software Engineering (MCP Tools)
│   ├── LandingPageGenerator-A2A        # Software Engineering (A2A Protocol)
│   ├── LandingPageGenerator-H_A2A      # Software Engineering (Heterogeneous A2A)
│   ├── LandingPageGenerator-MCP        # Software Engineering (MCP Tools)
│   ├── MarkdownValidator               # Software Engineering (Pure Framework)
│   ├── MarkdownValidator-MCP           # Software Engineering (MCP Tools)
│   ├── RecruitmentAssistant-A2A        # Communication (A2A Protocol)
│   ├── RecruitmentAssistant-H_A2A      # Communication (Heterogeneous A2A)
│   ├── RecruitmentAssistant-MCP        # Communication (MCP Tools)
│   ├── SocialMediaManager-A2A          # Content Generation (A2A Protocol)
│   ├── SocialMediaManager-H_A2A        # Content Generation (Heterogeneous A2A)
│   ├── SocialMediaManager-MCP          # Content Generation (MCP Tools)
│   ├── SQLAssistant-A2A                # Data Engineering (A2A Protocol)
│   ├── SQLAssistant-H_A2A              # Data Engineering (Heterogeneous A2A)
│   └── SQLAssistant-MCP                # Data Engineering (MCP Tools)