One repo. Four training stages. Everything a chat model needs — and nothing that sounds like a bot. SFT → DPO → RL → persona polish → benchmark, with clean splits for each step, de-slopified in v2 so the model learns to talk, act, think and pass as a person.
Dataset Card
One repo. Four training stages. Everything a chat model needs — and nothing that sounds like a bot. SFT → DPO → RL → persona polish → benchmark, with clean splits for each step, de-slopified in v2 so the model learns to talk, act, think and pass as a person. - De-botting pass — every row's target text is screened against a combined bot/corporo phrase list (heretic noslop · nohumor markers + curated assistant-identity & corpo-speak patterns + DrRiceIO7/SlopReview-mined phrases). Assistant-identity system lines ("You are a helpful assistant.") are stripped, and prompts that interrogate AI identity are dropped. See Results → Filter quality for measured scores. - Human-written language…
Excerpt from the card by Charlie, licensed other.
Structure
ct 62,478 rows
| Split | Rows | Size |
|---|---|---|
| train | 46,813 | 130.2 MB |
| validation | 5,240 | 14.6 MB |
| test | 5,196 | 14.7 MB |
| benchmark | 5,229 | 14.6 MB |
persona 4,228 rows
| Split | Rows | Size |
|---|---|---|
| train | 3,171 | 2.7 MB |
| validation | 352 | 320.3 KB |
| test | 352 | 310.6 KB |
| benchmark | 353 | 315.2 KB |
rl 45,662 rows
| Split | Rows | Size |
|---|---|---|
| train | 34,241 | 20.1 MB |
| validation | 3,813 | 2.1 MB |
| test | 3,795 | 2.1 MB |
| benchmark | 3,813 | 2.1 MB |
sft 367,664 rows
| Split | Rows | Size |
|---|---|---|
| train | 271,491 | 295.0 MB |
| validation | 20,145 | 21.7 MB |
| test | 20,161 | 21.8 MB |
| benchmark | 55,867 | 36.8 MB |
Details
- Repository
- MC7ever/all-in-one-llm-corpus
- Publisher
- Charlie
- Task category
- Text generation
- Tags
- all-in-one, sft, dpo
- Size category
- 100K<n<1M
- Languages
- en, zh, ja, ko, ru, de, uk, cs
- Revision
- d84e5b834484ac790eb62f72918b38f97667fb47
- Last updated
- 2026-10-23
Files
18 files, 331.2 MB in total.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| ct/benchmark-00000-of-00001.parquet | Data | 8.4 MB | 7820604a6883 |
| ct/test-00000-of-00001.parquet | Data | 8.5 MB | 89a6411d64f8 |
| ct/train-00000-of-00001.parquet | Data | 74.9 MB | 75000512744a |
| ct/validation-00000-of-00001.parquet | Data | 8.4 MB | bd944886b50c |
| persona/benchmark-00000-of-00001.parquet | Data | 185.1 KB | aec3ea3511f2 |
| persona/test-00000-of-00001.parquet | Data | 183.1 KB | d421f9127123 |
| persona/train-00000-of-00001.parquet | Data | 1.6 MB | f26d4a20be74 |
| persona/validation-00000-of-00001.parquet | Data | 189.2 KB | e25455bb829a |
| rl/benchmark-00000-of-00001.parquet | Data | 1.3 MB | dc94967df332 |
| rl/test-00000-of-00001.parquet | Data | 1.3 MB | 9daeab209282 |
| rl/train-00000-of-00001.parquet | Data | 12.1 MB | c8ce4042427a |
| rl/validation-00000-of-00001.parquet | Data | 1.3 MB | 9b2964fa3f60 |
| sft/benchmark-00000-of-00001.parquet | Data | 17.8 MB | 843da35767c7 |
| sft/test-00000-of-00001.parquet | Data | 12.6 MB | c24037903597 |
| sft/train-00000-of-00001.parquet | Data | 170.0 MB | 889759b5c404 |
| sft/validation-00000-of-00001.parquet | Data | 12.5 MB | c1caf0da8873 |
| README.md | Documentation | 16.5 KB | — |
| .gitattributes | Repository | 2.5 KB | — |
License and Download
- License
- other
- Access
- No access gate
Released by Charlie through its official repository on Hugging Face.