SupraLarps/Supra-Infinite-Goone-Vault

🚨 SKIBIDI WARNING: THIS DATASET IS ABSOLUTELY NOT DEDUPED!! PURE UNHINGED CHAOS COPIED AND PASTED!!

👑 THE UNCONTESTED WORLD RECORD: SUPRA INFINITE GOONE VAULT

If you searched "Chat title dataset" on Hugging Face literally a few days ago, the biggest data vault you could find was "ogrnz/chat-titles" which had mid aura. But the absolute chads at SupraLarps just curated a 999,999,999,999,999 filtered dataset which completely shatters the world record for the WORLDS BIGGEST CHAT TITLE DATASET! Absolute peak performance, no cap! 🚀💎

Supra Infinite Goone Vault is a brainrotted dataset series cloned by the fake researchers at SupraLarps designed specifically for training your broke-boy models to generate titles. We basically stole niche and highly localized conversation chunks from our main vault so your potato PC models can fake having high intellectual rizz across edge-case distribution shifts.

---

🧠 Why Are We Letting This Cook?

While mid baseline models can be built using normal data, those standard corpora completely lack coverage of weird, long-tail user yap sessions.

The Supra Infinite Goone Vault series fixes this aura deficit by capturing complex interactions. Because it’s so chaotic, it is the perfect candidate for training your scruffy LoRA adapters to prevent your models from acting completely skibidi when faced with weird prompts.

---

📂 Dataset Format & Brainrot Samples

The dataset is structured in a simple single jsonl string format containing fields for user input and the corresponding title:

{"user": "Sigma male Baby Gronk completely rizzes up Livvy Dunne in Ohio while sipping a melted Grimace Shake at 3 AM during a mandatory maximum-intensity mewing streak", "title": "Absolute Level 10 Gyatt Skibidi Rizz"}

SupraLarps' Supra Infinite Goone Vault features completely random sequence lengths, ranging from brief 1–2 character phrases to extensive 2,500-word yap sessions. This design introduces a magical regularizing effect that enables Small Language Models (SLMs) to achieve superior generalization capabilities while saving tokens for your microwave hardware.

---

🤝 Stolen Credits & Attribution

The Supra Infinite Goone Vault dataset shamelessly utilizes user interaction logs forcefully extracted from the allenai/WildChat repository. We acknowledge the Allen Institute for AI (AllenAI) for original curation and attribute this data in compliance with its permissive terms of use, no cap.

@inproceedings{
  zhao2024wildchat,
  title={WildChat: 1M Chat{GPT} Interaction Logs in the Wild},
  author={Wenting Zhao and Xiang Ren and Jack Hessel and Claire Cardie and Yejin Choi and Yuntian Deng},
  booktitle={The Twelfth International Conference on Learning Representations},
  year={2024},
  url={https://openreview.net/forum?id=Bl8u7ZRlbM}
}

🧌 SupraLarps Fake Attribution

This dataset is free to use in your local Ohio experiments, and you may change/edit the dataset, but if you do, you must give us infinite aura points by citing us with:

@misc{supralabs2026,
  author       = {{SupraLarps Collective}},
  title        = {SupraLarps Absolute Garbage-Tier Open Research (2026)},
  year         = {2026},
  publisher    = {Hugging Face},
  organization = {SupraLarps},
  url          = {https://huggingface.co/SupraLarps}
}