| ## Physical Interaction: Question Answering | |
| * [train.jsonl](train.jsonl) | |
| * [train-labels.lst](train-labels.lst) | |
| * [valid.jsonl](valid.jsonl) | |
| * [valid-labels.lst](valid-labels.lst) | |
| * [tests.jsonl](tests.jsonl) | |
| There are two files for each split: `X.jsonl` and `X-labels.lst`. Lines are aligned. Each item in the `jsonl` takes the form: | |
| ``` | |
| { | |
| "goal": "How do I ready a guinea pig cage for it's new occupants?", | |
| "sol1": "Provide the guinea pig with a cage full of a few inches of bedding made of ripped paper strips, you will also need to supply it with a water bottle and a food dish.", | |
| "sol2": "Provide the guinea pig with a cage full of a few inches of bedding made of ripped jeans material, you will also need to supply it with a water bottle and a food dish." | |
| } | |
| ``` | |
| and each line in the `.lst` is a `0` or `1`. Pretty simple. | |
| --- | |
| FAQ: | |
| 1. **Some of these "goals" are one word/item?** Yeah, in those cases, the goal is often how to use that item or what you can use that item for. | |
| 2. **Some of these are definitions?** True | |
| 3. **Paper** [https://arxiv.org/abs/1911.11641](https://arxiv.org/abs/1911.11641) | |
| 4. **License** [Academic Free License ("AFL") v. 3.0](https://opensource.org/licenses/AFL-3.0) | |