---
dataset_info:
  features:
  - name: text
    dtype: string
  - name: segments
    list:
    - name: label
      dtype: bool
    - name: text
      dtype: string
  - name: format
    dtype: string
  - name: n_explanations
    dtype: int64
  splits:
  - name: train
    num_bytes: 12553992020
    num_examples: 20000000
  download_size: 5406723121
  dataset_size: 12553992020
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---
# Othello (synthetic) (Llama-3 multi-format, input_output)

Continued-pre-training version of [`cfierro/othello-synthetic`](https://huggingface.co/datasets/cfierro/othello-synthetic), formatted for Axolotl's template-free `input_output` loss masking with Llama-3 BOS/EOS. Each game is written in one of several randomly-chosen move-sequence formats so the model is robust to how games are notated.

## Formats

Every row shares the same masked prefix `<|begin_of_text|>Game of bjk\n\n`; only the formatting of the move sequence (the trained body) changes. The default **fixed_prefix** (~70%) uses space-separated moves; six alternatives are each applied with ~5% probability. Black moves first. Othello lets a player with no legal move **pass**, so the colours do not strictly alternate: the player labels below come from the source dataset's `players` column (the true colour of each move), never from the move index.

All formats are `<|begin_of_text|>Game of bjk\n\n` + the body below + `<|end_of_text|>`:

```
fixed_prefix             f5 f4 c3 ...
player_1_2               Player 1: f5\nPlayer 2: f4\nPlayer 1: ...
black_white_colon        black: f5\nwhite: f4\nblack: ...
turn_n                   turn 1: f5\nturn 2: f4\nturn 3: ...
black_white_space        black f5\nwhite f4\nblack ...
comma                    f5, f4, c3, ...
paren_black_white_comma  (black) f5, (white) f4, (black) c3, ...
```

Stored under the `segments` schema consumed by `type: input_output`:

```json
{"segments": [
  {"label": false, "text": "<|begin_of_text|>Game of bjk\n\n"},
  {"label": true,  "text": "<formatted moves><|end_of_text|>"}
]}
```

- The masked prefix `<|begin_of_text|>Game of bjk\n\n` (`label: false`) is not in the loss but is still attended to as context.
- The formatted move sequence and the trailing `<|end_of_text|>` (`label: true`) are trained, mirroring a Llama-3 pre-training document terminator.
- Special tokens are written literally and consumed verbatim by Axolotl (it adds no BOS/EOS automatically). The chosen `format` and the original move string `text` are kept for inspection.

## Splits

Splits mirror the source dataset, [`cfierro/othello-synthetic`](https://huggingface.co/datasets/cfierro/othello-synthetic).

| Split | Rows |
|---|---:|
| `train` | 20,000,000 |

Train on part of it with HF split syntax, e.g. `train[:500000]`.

## Axolotl usage

```yaml
datasets:
  - path: cfierro/othello-llama3-multi-format
    type: input_output
    split: train[:500000]
train_on_inputs: false
```

## Provenance

- **Code**: generated with [`data/push_input_output_dataset.py` @ `8bc64a1c1e`](https://github.com/constanzafierro/rules-learning/blob/8bc64a1c1eca92528038b359fa15dfa6db753c30/data/push_input_output_dataset.py).
> ⚠️ The working tree had **uncommitted changes** when this dataset was generated, so the linked commit may not exactly match the code that ran. Commit the generator before generating for a faithful link.

- **Built**: 2026-08-06T08:49:17
- **Command**:

```bash
python data/push_input_output_dataset.py --game othello --source_splits train --repo cfierro/othello-llama3-multi-format
```

- **Resolved arguments**:

```json
{
  "game": "othello",
  "source_repo": "cfierro/othello-synthetic",
  "num_proc": null,
  "source_splits": [
    "train"
  ],
  "repo": "cfierro/othello-llama3-multi-format",
  "model_family": "llama3",
  "seed": 0,
  "alt_prob": 0.05,
  "mode": "plain",
  "explain_game_ratio": 0.5,
  "explain_count_dist": "0.4,0.35,0.2,0.05",
  "explanation_templates": null,
  "explanation_templates_split": "train",
  "free_adj_ratio": 0.8,
  "dry_run": false,
  "private": false,
  "inspect": false,
  "tokenizer": null
}
```