| --- |
| license: mit |
| task_categories: |
| - text-generation |
| tags: |
| - philosophy |
| - lesswrong |
| - long-sequences |
| - posts |
| size_categories: |
| - 10K<n<100K |
| --- |
| |
| # What is LessWrong? |
|
|
| LessWrong is a community blog and forum dedicated to improving human reasoning and decision-making, aiming to help people hold more accurate beliefs and be more effective, or "less wrong," daily. |
| This dataset contains over twenty-six thousand posts from 2009 and onward. |
|
|
| ## Stats |
|
|
| | Key | Value | |
| |----------------------|----------------| |
| | Entries | 26,517 | |
| | Total Tokens (GPT2) | 96,399,665 | |
| | Total Words | 53,966,975 | |
| | Avg Tokens / Entry | 3,635.39 | |
| | Avg Words / Entry | 2,035.18 | |
|
|
| We counted the tokens using [GPT2's](https://huggingface.co/openai-community/gpt2) tokenizer. |
|
|
| --- |
|
|
| ## Use Cases |
|
|
| Pretraining and fine-tuning language models on long-form rationalist and AI-safety discourse. |
