# AI-NativeBench Data

This `data/` directory contains the dataset for **"AI-NativeBench: An Open-Source White-Box Agentic Benchmark Suite for AI-Native Systems"**.

It is split into two parts:

- `raw/`: original, unmodified experimental data organized by model and task/architecture.
- `processed/`: derived run artifacts, aggregated tables, figures, and analysis scripts built on top of the raw traces.

## Directory overview

```text
data/
├── raw/         # Raw experimental data (per-model, per task/architecture)
└── processed/   # Processed results, aggregation outputs, figures, and scripts
```

## `raw/` (raw experimental data)

- Organized hierarchically by **Model → Application/Architecture**.
- Provides a white-box view of executions, including distributed tracing data.

See `raw/README.md` for the full model list and task/architecture folder naming.

## `processed/` (results & analysis artifacts)

- Contains per-run outputs (e.g., `langfuse_trace.json`, extracted `execution_path.md`, run `metadata.json`, and task outputs).
- Includes RQ-level analysis artifacts under `RQ1/`, `RQ2/`, `RQ3/` (CSVs, PDFs, and plotting/aggregation scripts).

See `processed/README.md` for the detailed layout and reproduction entry points.