agentlans/llm-prompt-collection

---
configs:
- config_name: all
data_files:
- path:
- prompts.jsonl.zst
split: train
- config_name: prompts_k100
data_files:
- path:
- prompts_k100.jsonl.zst
split: train
- config_name: prompts_k1000
data_files:
- path:
- prompts_k1000.jsonl.zst
split: train
- config_name: prompts_k10000
data_files:
- path:
- prompts_k10000.jsonl.zst
split: train
- config_name: prompts_k100000
data_files:
- path:
- prompts_k100000.jsonl.zst
split: train
- config_name: prompts_k200
data_files:
- path:
- prompts_k200.jsonl.zst
split: train
- config_name: prompts_k2000
data_files:
- path:
- prompts_k2000.jsonl.zst
split: train
- config_name: prompts_k20000
data_files:
- path:
- prompts_k20000.jsonl.zst
split: train
- config_name: prompts_k200000
data_files:
- path:
- prompts_k200000.jsonl.zst
split: train
- config_name: prompts_k500
data_files:
- path:
- prompts_k500.jsonl.zst
split: train
- config_name: prompts_k5000
data_files:
- path:
- prompts_k5000.jsonl.zst
split: train
- config_name: prompts_k50000
data_files:
- path:
- prompts_k50000.jsonl.zst
split: train
- config_name: prompts_k500000
data_files:
- path:
- prompts_k500000.jsonl.zst
split: train
- config_name: screened_prompts
data_files:
- path:
- screened_prompts.jsonl.zst
split: train
default: true
language:
- en
- multilingual
---
# LLM Prompt Collection

- Prompts from the first 100 000 rows of each dataset were collected then deduplicated and shuffled.
- The `prompts_k*` configs are semantically clustered subsets of the `all` config for diversity and coverage.
- The `screened_prompts` config is a subset of safe, high-quality prompts for the `all` config as classified using [agentlans/bge-small-en-v1.5-prompt-screener](https://huggingface.co/agentlans/bge-small-en-v1.5-prompt-screener)

| Source | Rows |
|:---|---:|
| [agentlans/chatgpt all](https://huggingface.co/datasets/agentlans/chatgpt/viewer/all) | 100 000 |
| [agentlans/magpie all](https://huggingface.co/datasets/agentlans/magpie/viewer/all) | 99 982 |
| [agentlans/epic-thinking](https://huggingface.co/datasets/agentlans/epic-thinking) | 98 926 |
| [agentlans/Locutusque-hercules-v6.9](https://huggingface.co/datasets/agentlans/Locutusque-hercules-v6.9) | 97 115 |
| [agentlans/rombodawg-Everything_Instruct](https://huggingface.co/datasets/agentlans/rombodawg-Everything_Instruct) | 96 502 |
| [agentlans/claude all](https://huggingface.co/datasets/agentlans/claude/viewer/all) | 96 406 |
| [agentlans/thomas-yanxin-MT-SFT-ShareGPT](https://huggingface.co/datasets/agentlans/thomas-yanxin-MT-SFT-ShareGPT) | 94 143 |
| [agentlans/NousResearch-Hermes-3-Dataset](https://huggingface.co/datasets/agentlans/NousResearch-Hermes-3-Dataset) | 94 097 |
| [agentlans/QuixiAI-dolphin-distill](https://huggingface.co/datasets/agentlans/QuixiAI-dolphin-distill) | 87 897 |