👁 Website
👁 GitHub
👁 Hugging Face
👁 Follow on X

 _ _ _ _______ ___________ _ _ ___________ _ ______ 
| | | | \ | | _ \ ___| ___ \ | | || _ | ___ \ | | _ \
| | | | \| | | | | |__ | |_/ / | | || | | | |_/ / | | | | |
| | | | . ` | | | | __|| /| |/\| || | | | /| | | | | |
| |_| | |\ | |/ /| |___| |\ \\ /\ /\ \_/ / |\ \| |___| |/ / 
 \___/\_| \_/___/ \____/\_| \_|\/ \/ \___/\_| \_\_____/___/ 
 

UNDERWORLD Dataset v3

• Visualizes the Fast Inverse Square Root (FISR / Quake III) algorithm.
• 120 rows total (train split only).
• Main content: 1280px PNG image frames showing bit hacks, Newton-Raphson steps, error surfaces, 3D math plots.
• Numerical_data.csv (regression), metadata.json (conditional).
• Size ~3.5 MB. Generated via UNDERGROUND: FISR tool (downloadable in repo).
• Magic Number: 0x5f23aac5
• Newton Iterations: 3
• Input Range: 0.1 to 1000
• Maximum Error: 1.1742636926798086e+287%

Generated with UNDERWORLD: FISR by webXOS, Educational visualization of the Quake III Arena optimization algorithm.

The UNDERWORLD app by webXOS is available for download in the /underworld/ folder of this repo so users can create their own datasets.

Use cases:

• Training ML models for fault detection / anomaly detection in time-series or sensor data.

• Simulating hardware faults (bit flips, stuck-at, etc.) for robust AI / embedded ML.

• Reliability engineering: predict system failures under errors.

• Synthetic data for safety-critical systems (automotive, aerospace, IoT) where real fault data is rare.

• Benchmarking error-correction / resilient algorithms.

• Visual sequence learning → train models on math visualization sequences (frame prediction, video understanding).

• Image-to-text / captioning → describe FISR steps from images.

• Visual question answering → QA on algorithm visuals.

• Regression from images → predict error metrics from visualization frames.

• Educational multimodal models → teach bit manipulation / fast math approx.

• Conditional generation → use metadata to condition on input range/error.

• 3D math function visualization benchmark → compare rendering / understanding.

Education:

1. The Fast Inverse Square Root algorithm implementation
2. Error analysis of the approximation
3. 3D visualization of mathematical functions
4. Bit-level manipulation techniques

Usage for Training:

1. Use frames/ for visual sequence learning
2. Use numerical_data.csv for regression tasks
3. Use metadata.json for conditional generation
4. Train models to understand optimization algorithms

Citation:

If you use this dataset, please cite:

webXOS

License:

MIT