WIA Datasets
Collection
World-model datasets for the WIA project. • 4 items • Updated • 1
Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Job has been terminated due to a temporary spike in resource usage and may be restarted later.
Error code: JobManagerCrashedError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
messages list | source string | original_source string | original_id string | metadata dict | original_tokens int64 | sequence_field string | fragment_index int64 | fragment_start_pair int64 | fragment_end_pair int64 | system_prompt_variant_index int64 |
|---|---|---|---|---|---|---|---|---|---|---|
[
{
"role": "system",
"content": "You are a helpful general agent trained on ADP standard trajectories.\n\nYou may call tools to solve the user's task. Tool calls must be written directly in assistant content using the following XML format:\n\n<tool_call>\n<function=tool_name>\n<parameter=parameter_name>\nJSO... | ADP-GeneralAgent | swe-smith | django__channels.a144b4b8.pr_2002.smujyx7z_1 | {
"details": {}
} | null | null | null | null | null | null |
[
{
"role": "system",
"content": "You are a helpful general agent trained on ADP standard trajectories.\n\nYou may call tools to solve the user's task. Tool calls must be written directly in assistant content using the following XML format:\n\n<tool_call>\n<function=tool_name>\n<parameter=parameter_name>\nJSO... | ADP-GeneralAgent | toucan_1_5m | toucan_sample_1152173 | {
"details": {
"available_apis": "def open_library_mcp_server_get_book_by_title(title: str) -> dict:\n \"\"\"Search for a book by its title on Open Library.\n\n Args:\n ----\n title: The title of the book to search for.\n\n \"\"\"\n pass\n\n\ndef price_mcp_server_Steam Games Price Checker(ap... | null | null | null | null | null | null |
[
{
"role": "system",
"content": "You are a helpful general agent trained on ADP standard trajectories.\n\nYou may call tools to solve the user's task. Tool calls must be written directly in assistant content using the following XML format:\n\n<tool_call>\n<function=tool_name>\n<parameter=parameter_name>\nJSO... | ADP-GeneralAgent | orca_agentinstruct | orca_open_domain_qa67926 | {
"details": {}
} | null | null | null | null | null | null |
[
{
"role": "system",
"content": "You are an Android app state simulator.\n\nThe user message provides a mobile UI trajectory context and one current action. Predict the observation emitted immediately after that action.\n\nThe observation is a textual tree of mobile accessibility data. Depending on the sourc... | MobileViews | null | null | null | null | null | null | null | null | null |
[
{
"role": "system",
"content": "You are a helpful general agent trained on ADP standard trajectories.\n\nYou may call tools to solve the user's task. Tool calls must be written directly in assistant content using the following XML format:\n\n<tool_call>\n<function=tool_name>\n<parameter=parameter_name>\nJSO... | ADP-GeneralAgent | toucan_1_5m | toucan_sample_108525 | {
"details": {
"available_apis": "def language_detection_server_detect(text: str) -> dict:\n \"\"\"\n MCP Tool: Detect language of the input text.\n \n\n Args:\n ----\n text: \n\n \"\"\"\n pass\n\n\ndef clear_thought_sequentialthinking(thought: str, thoughtNumber: int, totalThoughts: i... | null | null | null | null | null | null |
[
{
"role": "system",
"content": "You are a helpful assistant that can interact with a computer to solve software engineering tasks.\n\nYou may call tools to inspect, modify, test the repository, or submit an answer. Tool calls must be written directly in assistant content using the following XML format:\n\n<... | GeneralAgent | skill_medium_dependency_management | episode-7 | null | 12,545 | conversations | null | null | null | null |
[{"role":"system","content":"You are a web browser world model.\n\nGiven a trajectory with an initia(...TRUNCATED) | WebWorld | null | null | null | null | null | null | null | null | null |
[{"role":"system","content":"You are a helpful general agent trained on ADP standard trajectories.\n(...TRUNCATED) | ADP-GeneralAgent | toucan_1_5m | toucan_sample_49409 | {"details":{"available_apis":"def 12306_mcp_server_search(date: str, fromCity: str, toCity: str) -> (...TRUNCATED) | null | null | null | null | null | null |
[{"role":"system","content":"You are a research browser world model.\n\nGiven a global research goal(...TRUNCATED) | OpenResearcher | null | null | null | null | null | null | null | null | null |
[{"role":"system","content":"You are a helpful general agent trained on ADP standard trajectories.\n(...TRUNCATED) | ADP-GeneralAgent | toucan_1_5m | toucan_sample_1312065 | {"details":{"available_apis":"def coin_price_fetcher_getTokenPrice(token: str) -> dict:\n \"\"\"g(...TRUNCATED) | null | null | null | null | null | null |
End of preview.
WIA-Corpus-CPT
File Purpose and Sample Count
Each line is one sample. The following 16 files jointly form the train
dataset for continued pretraining of world models, with 5,032,034 samples
in total.
| File | Purpose | Samples | Top-level fields |
|---|---|---|---|
cpt-data-00000-of-00016.jsonl |
Continued-pretraining shard 00 | 314,503 | 11 types; 2 required and 9 optional |
cpt-data-00001-of-00016.jsonl |
Continued-pretraining shard 01 | 314,503 | Same as above |
cpt-data-00002-of-00016.jsonl |
Continued-pretraining shard 02 | 314,502 | Same as above |
cpt-data-00003-of-00016.jsonl |
Continued-pretraining shard 03 | 314,502 | Same as above |
cpt-data-00004-of-00016.jsonl |
Continued-pretraining shard 04 | 314,502 | Same as above |
cpt-data-00005-of-00016.jsonl |
Continued-pretraining shard 05 | 314,502 | Same as above |
cpt-data-00006-of-00016.jsonl |
Continued-pretraining shard 06 | 314,502 | Same as above |
cpt-data-00007-of-00016.jsonl |
Continued-pretraining shard 07 | 314,502 | Same as above |
cpt-data-00008-of-00016.jsonl |
Continued-pretraining shard 08 | 314,502 | Same as above |
cpt-data-00009-of-00016.jsonl |
Continued-pretraining shard 09 | 314,502 | Same as above |
cpt-data-00010-of-00016.jsonl |
Continued-pretraining shard 10 | 314,502 | Same as above |
cpt-data-00011-of-00016.jsonl |
Continued-pretraining shard 11 | 314,502 | Same as above |
cpt-data-00012-of-00016.jsonl |
Continued-pretraining shard 12 | 314,502 | Same as above |
cpt-data-00013-of-00016.jsonl |
Continued-pretraining shard 13 | 314,502 | Same as above |
cpt-data-00014-of-00016.jsonl |
Continued-pretraining shard 14 | 314,502 | Same as above |
cpt-data-00015-of-00016.jsonl |
Continued-pretraining shard 15 | 314,502 | Same as above |
Field Definitions
| Field | Type | Required | Purpose |
|---|---|---|---|
messages |
array | yes | Ordered conversation messages, actions, and environment observations. |
source |
string | yes | Identifies the sample source. |
original_source |
string | no | Name of the original source. |
original_id |
string | no | Identifier of the original sample. |
metadata |
object | no | Additional source information stored in metadata.details; subfields vary by source. |
original_tokens |
integer | no | Token count recorded for the original sample. |
sequence_field |
string | no | Field name used to extract the sequence from the original data. |
fragment_index |
integer | no | Index of the trajectory fragment. |
fragment_start_pair |
integer | no | Position of the first message pair in the fragment. |
fragment_end_pair |
integer | no | Position of the last message pair in the fragment. |
system_prompt_variant_index |
integer | no | Index of the system-prompt variant. |
Each message in messages[] contains three fields:
| Field | Type | Purpose |
|---|---|---|
role |
string | Message role. |
content |
string | Message content. |
loss |
boolean | Whether the message contributes to the training loss. |
Download
pip install -U huggingface_hub
from huggingface_hub import snapshot_download
snapshot_download(
repo_id="LightningCreeper/WIA-Corpus-CPT",
repo_type="dataset",
allow_patterns=["*.jsonl"],
local_dir="WIA-Corpus-CPT",
)
- Downloads last month
- 109