---
configs:
- config_name: default
  default: true
  data_files:
  - split: train
    path: data/train/**/*.parquet
  - split: validation
    path: data/validation/**/*.parquet
  - split: test
    path: data/test/**/*.parquet
- config_name: ABCD
  data_files:
  - split: train
    path:
    - data/train/ABCD.parquet
  - split: validation
    path:
    - data/validation/ABCD.parquet
  - split: test
    path:
    - data/test/ABCD.parquet
- config_name: AMI
  data_files:
  - split: train
    path:
    - data/train/AMI.parquet
  - split: validation
    path:
    - data/validation/AMI.parquet
  - split: test
    path:
    - data/test/AMI.parquet
- config_name: AirDialogue
  data_files:
  - split: train
    path:
    - data/train/AirDialogue.parquet
  - split: validation
    path:
    - data/validation/AirDialogue.parquet
  - split: test
    path:
    - data/test/AirDialogue.parquet
- config_name: BiTOD
  data_files:
  - split: train
    path:
    - data/train/BiTOD.parquet
  - split: validation
    path:
    - data/validation/BiTOD.parquet
  - split: test
    path:
    - data/test/BiTOD.parquet
- config_name: CCPE-M
  data_files:
  - split: train
    path:
    - data/train/CCPE-M.parquet
  - split: validation
    path:
    - data/validation/CCPE-M.parquet
  - split: test
    path:
    - data/test/CCPE-M.parquet
- config_name: CRD3
  data_files:
  - split: train
    path:
    - data/train/CRD3.parquet
  - split: validation
    path:
    - data/validation/CRD3.parquet
  - split: test
    path:
    - data/test/CRD3.parquet
- config_name: CaSiNo
  data_files:
  - split: train
    path:
    - data/train/CaSiNo.parquet
  - split: validation
    path:
    - data/validation/CaSiNo.parquet
  - split: test
    path:
    - data/test/CaSiNo.parquet
- config_name: CraigslistBargains
  data_files:
  - split: train
    path:
    - data/train/CraigslistBargains.parquet
  - split: validation
    path:
    - data/validation/CraigslistBargains.parquet
  - split: test
    path:
    - data/test/CraigslistBargains.parquet
- config_name: Deal or No Deal
  data_files:
  - split: train
    path:
    - data/train/Deal or No Deal.parquet
  - split: validation
    path:
    - data/validation/Deal or No Deal.parquet
  - split: test
    path:
    - data/test/Deal or No Deal.parquet
- config_name: DuRecDial-2.0
  data_files:
  - split: train
    path:
    - data/train/DuRecDial-2.0.parquet
  - split: validation
    path:
    - data/validation/DuRecDial-2.0.parquet
  - split: test
    path:
    - data/test/DuRecDial-2.0.parquet
- config_name: FITS
  data_files:
  - split: train
    path:
    - data/train/FITS.parquet
  - split: validation
    path:
    - data/validation/FITS.parquet
  - split: test
    path:
    - data/test/FITS.parquet
- config_name: FRAMES
  data_files:
  - split: train
    path:
    - data/train/FRAMES.parquet
  - split: validation
    path:
    - data/validation/FRAMES.parquet
  - split: test
    path:
    - data/test/FRAMES.parquet
- config_name: Humanual-Book
  data_files:
  - split: train
    path:
    - data/train/Humanual-Book.parquet
  - split: validation
    path:
    - data/validation/Humanual-Book.parquet
  - split: test
    path:
    - data/test/Humanual-Book.parquet
- config_name: ICSI
  data_files:
  - split: train
    path:
    - data/train/ICSI.parquet
  - split: validation
    path:
    - data/validation/ICSI.parquet
  - split: test
    path:
    - data/test/ICSI.parquet
- config_name: INSPIRED
  data_files:
  - split: train
    path:
    - data/train/INSPIRED.parquet
  - split: validation
    path:
    - data/validation/INSPIRED.parquet
  - split: test
    path:
    - data/test/INSPIRED.parquet
- config_name: KVRET
  data_files:
  - split: train
    path:
    - data/train/KVRET.parquet
  - split: validation
    path:
    - data/validation/KVRET.parquet
  - split: test
    path:
    - data/test/KVRET.parquet
- config_name: MSC
  data_files:
  - split: train
    path:
    - data/train/MSC.parquet
  - split: validation
    path:
    - data/validation/MSC.parquet
  - split: test
    path:
    - data/test/MSC.parquet
- config_name: MULTIWOZ2_2
  data_files:
  - split: train
    path:
    - data/train/MULTIWOZ2_2.parquet
  - split: validation
    path:
    - data/validation/MULTIWOZ2_2.parquet
  - split: test
    path:
    - data/test/MULTIWOZ2_2.parquet
- config_name: MetaLWOZ
  data_files:
  - split: train
    path:
    - data/train/MetaLWOZ.parquet
  - split: validation
    path:
    - data/validation/MetaLWOZ.parquet
  - split: test
    path:
    - data/test/MetaLWOZ.parquet
- config_name: MuDoCo
  data_files:
  - split: train
    path:
    - data/train/MuDoCo.parquet
  - split: validation
    path:
    - data/validation/MuDoCo.parquet
  - split: test
    path:
    - data/test/MuDoCo.parquet
- config_name: MulDoGO
  data_files:
  - split: train
    path:
    - data/train/MulDoGO.parquet
  - split: validation
    path:
    - data/validation/MulDoGO.parquet
  - split: test
    path:
    - data/test/MulDoGO.parquet
- config_name: MutualFriends
  data_files:
  - split: train
    path:
    - data/train/MutualFriends.parquet
  - split: validation
    path:
    - data/validation/MutualFriends.parquet
  - split: test
    path:
    - data/test/MutualFriends.parquet
- config_name: OpenDialKG
  data_files:
  - split: train
    path:
    - data/train/OpenDialKG.parquet
  - split: validation
    path:
    - data/validation/OpenDialKG.parquet
  - split: test
    path:
    - data/test/OpenDialKG.parquet
- config_name: Redial
  data_files:
  - split: train
    path:
    - data/train/Redial.parquet
  - split: validation
    path:
    - data/validation/Redial.parquet
  - split: test
    path:
    - data/test/Redial.parquet
- config_name: SGD
  data_files:
  - split: train
    path:
    - data/train/SGD.parquet
  - split: validation
    path:
    - data/validation/SGD.parquet
  - split: test
    path:
    - data/test/SGD.parquet
- config_name: STAR
  data_files:
  - split: train
    path:
    - data/train/STAR.parquet
  - split: validation
    path:
    - data/validation/STAR.parquet
  - split: test
    path:
    - data/test/STAR.parquet
- config_name: Taskmaster-4
  data_files:
  - split: train
    path:
    - data/train/Taskmaster-4.parquet
  - split: validation
    path:
    - data/validation/Taskmaster-4.parquet
  - split: test
    path:
    - data/test/Taskmaster-4.parquet
- config_name: Taskmaster1
  data_files:
  - split: train
    path:
    - data/train/Taskmaster1.parquet
  - split: validation
    path:
    - data/validation/Taskmaster1.parquet
  - split: test
    path:
    - data/test/Taskmaster1.parquet
- config_name: Taskmaster2
  data_files:
  - split: train
    path:
    - data/train/Taskmaster2.parquet
  - split: validation
    path:
    - data/validation/Taskmaster2.parquet
  - split: test
    path:
    - data/test/Taskmaster2.parquet
- config_name: Taskmaster3
  data_files:
  - split: train
    path:
    - data/train/Taskmaster3.parquet
  - split: validation
    path:
    - data/validation/Taskmaster3.parquet
  - split: test
    path:
    - data/test/Taskmaster3.parquet
- config_name: Topical-Chat
  data_files:
  - split: train
    path:
    - data/train/Topical-Chat.parquet
  - split: validation
    path:
    - data/validation/Topical-Chat.parquet
  - split: test
    path:
    - data/test/Topical-Chat.parquet
- config_name: TweetSumm
  data_files:
  - split: train
    path:
    - data/train/TweetSumm.parquet
  - split: validation
    path:
    - data/validation/TweetSumm.parquet
  - split: test
    path:
    - data/test/TweetSumm.parquet
- config_name: chitchat-dataset
  data_files:
  - split: train
    path:
    - data/train/chitchat-dataset.parquet
  - split: validation
    path:
    - data/validation/chitchat-dataset.parquet
  - split: test
    path:
    - data/test/chitchat-dataset.parquet
- config_name: contextual-abuse
  data_files:
  - split: train
    path:
    - data/train/convokit/contextual-abuse.parquet
  - split: validation
    path:
    - data/validation/convokit/contextual-abuse.parquet
  - split: test
    path:
    - data/test/convokit/contextual-abuse.parquet
- config_name: deli-corpus
  data_files:
  - split: train
    path:
    - data/train/convokit/deli-corpus.parquet
  - split: validation
    path:
    - data/validation/convokit/deli-corpus.parquet
  - split: test
    path:
    - data/test/convokit/deli-corpus.parquet
- config_name: diplomacy-corpus
  data_files:
  - split: train
    path:
    - data/train/convokit/diplomacy-corpus.parquet
  - split: validation
    path:
    - data/validation/convokit/diplomacy-corpus.parquet
  - split: test
    path:
    - data/test/convokit/diplomacy-corpus.parquet
- config_name: fomc-corpus
  data_files:
  - split: train
    path:
    - data/train/convokit/fomc-corpus.parquet
  - split: validation
    path:
    - data/validation/convokit/fomc-corpus.parquet
  - split: test
    path:
    - data/test/convokit/fomc-corpus.parquet
- config_name: gap-corpus
  data_files:
  - split: train
    path:
    - data/train/convokit/gap-corpus.parquet
  - split: validation
    path:
    - data/validation/convokit/gap-corpus.parquet
  - split: test
    path:
    - data/test/convokit/gap-corpus.parquet
- config_name: iq2-corpus
  data_files:
  - split: train
    path:
    - data/train/convokit/iq2-corpus.parquet
  - split: validation
    path:
    - data/validation/convokit/iq2-corpus.parquet
  - split: test
    path:
    - data/test/convokit/iq2-corpus.parquet
- config_name: llm-facilitation-corpus
  data_files:
  - split: train
    path:
    - data/train/convokit/llm-facilitation-corpus.parquet
  - split: validation
    path:
    - data/validation/convokit/llm-facilitation-corpus.parquet
  - split: test
    path:
    - data/test/convokit/llm-facilitation-corpus.parquet
- config_name: news-interview
  data_files:
  - split: train
    path:
    - data/train/convokit/news-interview.parquet
  - split: validation
    path:
    - data/validation/convokit/news-interview.parquet
  - split: test
    path:
    - data/test/convokit/news-interview.parquet
- config_name: spolin-corpus
  data_files:
  - split: train
    path:
    - data/train/convokit/spolin-corpus.parquet
  - split: validation
    path:
    - data/validation/convokit/spolin-corpus.parquet
  - split: test
    path:
    - data/test/convokit/spolin-corpus.parquet
- config_name: wizard_of_internet
  data_files:
  - split: train
    path:
    - data/train/wizard_of_internet.parquet
  - split: validation
    path:
    - data/validation/wizard_of_internet.parquet
  - split: test
    path:
    - data/test/wizard_of_internet.parquet
- config_name: wizard_of_wikipedia
  data_files:
  - split: train
    path:
    - data/train/wizard_of_wikipedia.parquet
  - split: validation
    path:
    - data/validation/wizard_of_wikipedia.parquet
  - split: test
    path:
    - data/test/wizard_of_wikipedia.parquet
---

# Human user-simulation mid-training

A dataset for training models to simulate human conversational behavior.

## Schema

| Column | Type | Meaning |
| --- | --- | --- |
| `conversation_id` | string | Conversation/branch/target perspective ID; original IDs are retained in local lineage. |
| `dataset` | string | Original source dataset name. |
| `category` | string | Behavioral category. |
| `messages` | list of structs | Full ordered conversation; each message has string `role` and `content` fields. |

Roles are already oriented for user-simulator training:

- `system`: optional persona, goal, and simulation instructions.
- `user`: the other conversational participant.
- `assistant`: the real human whose behavior is being modeled; these are the training targets.

## Documentation example

```json
{
  "conversation_id": "illustration-1",
  "dataset": "illustration-only",
  "category": "illustrative-category",
  "messages": [
    {"role": "system", "content": "You are a student asking for an explanation."},
    {"role": "user", "content": "Which step would you like to discuss?"},
    {"role": "assistant", "content": "Why do we divide both sides by three?"}
  ]
}
```

<!-- available-datasets:start -->
## Available Datasets

**Dataset count: 44**

Select an individual dataset by its name below, or use `default` for the combined collection.

```python
from datasets import load_dataset

dataset = load_dataset("ajayvikram/user-simulation-midtraining", "ABCD")
```

Datasets are grouped by behavioral category:

```python
{
  "Collaboration & Coordination": [
    "AMI", "ICSI", "MutualFriends", "deli-corpus", "fomc-corpus", "gap-corpus",
    "llm-facilitation-corpus", "spolin-corpus"
  ],
  "General Conversation & Social Interaction": [
    "MSC", "Topical-Chat", "chitchat-dataset"
  ],
  "Information Seeking & Clarification": [
    "FITS", "news-interview", "wizard_of_internet", "wizard_of_wikipedia"
  ],
  "Negotiation, Persuasion & Debate": [
    "CaSiNo", "CraigslistBargains", "Deal or No Deal", "contextual-abuse", "diplomacy-corpus",
    "iq2-corpus"
  ],
  "Preferences & Recommendations": [
    "CCPE-M", "DuRecDial-2.0", "Humanual-Book", "INSPIRED", "OpenDialKG", "Redial"
  ],
  "Role-Playing & Character Simulation": [
    "CRD3"
  ],
  "Task-Oriented Assistance & Services": [
    "ABCD", "AirDialogue", "BiTOD", "FRAMES", "KVRET", "MULTIWOZ2_2", "MetaLWOZ", "MuDoCo",
    "MulDoGO", "SGD", "STAR", "Taskmaster-4", "Taskmaster1", "Taskmaster2", "Taskmaster3",
    "TweetSumm"
  ]
}
```
<!-- available-datasets:end -->