Datasets:
entry_id int64 0 6.68k | shard large_stringclasses 1
value | episode_id large_stringlengths 64 64 | source_id large_stringclasses 1
value | stream large_stringclasses 1
value | split large_stringclasses 3
values | family_id large_stringlengths 72 72 | tokens int64 11 6.93k | scored_positions int64 6 664 | outcome large_stringclasses 1
value | verified_success bool 1
class | messages listlengths 1 24 | tools large_stringclasses 1
value | episode_json large_stringlengths 2.45k 74.8k |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0 | episodes-000000.jsonl | 4155320c4bc3f22b427376e97bee7068d47dd70ba1bd8473f8d93031ff609e47 | simia-agentbench-sft | environment_trajectories | train | content:e35b256f0613cb859fd2b3d2a92472370c05c5d3ba25864c8f82ea94543fb78f | 371 | 38 | unknown | false | [
{
"role": "user",
"text": "'''\n<html>\n<div>\n<a id=0>\n<span> Italian Bistro </span>\n<img available-dinner />\n<img available-lunch />\n</a>\n<a id=1>\n<span> Sushi Place </span>\n<img available-dinner />\n</a>\n<a id=2>\n<span> Vegan Cafe </span>\n<img available-breakfast />\n<img available-lunch />\n</... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"4155320c4bc3f22b427376e97bee7068d47dd70ba1bd8473f8d93031ff609e47","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:e35b256f0613cb859fd2b3d2a92472370c05c5d3ba25864c8f82ea94543fb78f","frozen_split":null,"parent_episo... |
1 | episodes-000000.jsonl | 77544074aa11843a6605b7e94608a22580bcdf67e7c6a193a00797f45cf4febe | simia-agentbench-sft | environment_trajectories | train | content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e | 772 | 151 | unknown | false | [
{
"role": "user",
"text": "You are web shopping.\nI will give you instructions about what to do.\nYou have to follow the instructions.\nEvery round I will give you an observation and a list of available actions, you have to respond an action based on the state and instruction.\nYou can use search action if ... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"77544074aa11843a6605b7e94608a22580bcdf67e7c6a193a00797f45cf4febe","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e","frozen_split":null,"parent_episo... |
2 | episodes-000000.jsonl | 9a66bf1a6aa023b8723c8bdd2a423b90d7843caec0cc3c7158b66d8210fa6773 | simia-agentbench-sft | environment_trajectories | train | content:9b7000ecc4ab8a6f0d2f16e3212b7a945c9fb152dc391c34ff71efac94ad2a68 | 351 | 27 | unknown | false | [
{
"role": "user",
"text": "'''\n<html> <div main> <nav primary> <a id=0> Home </a> <a id=1> Electronics </a> <a id=2> Fashion </a> </nav> <section secondary> <a id=3> Sort by Price </a> <a id=4> Filter: On Sale </a> <div class=\"items\"> <span class=\"item\" data-id=5> Laptop - $999 </span> <span class=\"it... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"9a66bf1a6aa023b8723c8bdd2a423b90d7843caec0cc3c7158b66d8210fa6773","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:9b7000ecc4ab8a6f0d2f16e3212b7a945c9fb152dc391c34ff71efac94ad2a68","frozen_split":null,"parent_episo... |
3 | episodes-000000.jsonl | b3bddb344b2f4418a019cc6f8cb2c290858cf9c11607a837876ba3d491645b24 | simia-agentbench-sft | environment_trajectories | train | content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e | 720 | 136 | unknown | false | [
{
"role": "user",
"text": "You are web shopping.\nI will give you instructions about what to do.\nYou have to follow the instructions.\nEvery round I will give you an observation and a list of available actions, you have to respond an action based on the state and instruction.\nYou can use search action if ... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"b3bddb344b2f4418a019cc6f8cb2c290858cf9c11607a837876ba3d491645b24","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e","frozen_split":null,"parent_episo... |
4 | episodes-000000.jsonl | 94a8325562bbaacbdd6b704dfd55988283c3cec554ff3c51a8f7fb4de0021d99 | simia-agentbench-sft | environment_trajectories | train | content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e | 814 | 161 | unknown | false | [
{
"role": "user",
"text": "You are web shopping.\nI will give you instructions about what to do.\nYou have to follow the instructions.\nEvery round I will give you an observation and a list of available actions, you have to respond an action based on the state and instruction.\nYou can use search action if ... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"94a8325562bbaacbdd6b704dfd55988283c3cec554ff3c51a8f7fb4de0021d99","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e","frozen_split":null,"parent_episo... |
5 | episodes-000000.jsonl | f95ddaff0c88dfafacd5846ad0b1bab4563b6511dee34c746db9597c2e1618ba | simia-agentbench-sft | environment_trajectories | train | content:892c83914d23f937cf048bc9970e40c9ca7386af70727dfe9cb28673410b60e2 | 501 | 91 | unknown | false | [
{
"role": "user",
"text": "'''\n<html>\n<div>\n<nav>\n<a id=0>Flights</a>\n<a id=1>Hotels</a>\n<a id=2>Cars</a>\n</nav>\n<iframe id=\"searchFlights\">\n<div>\n<input id=10 placeholder=\"Origin\">\n<input id=11 placeholder=\"Destination\">\n<input id=12 placeholder=\"Departure Date\">\n<button id=13>Search F... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"f95ddaff0c88dfafacd5846ad0b1bab4563b6511dee34c746db9597c2e1618ba","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:892c83914d23f937cf048bc9970e40c9ca7386af70727dfe9cb28673410b60e2","frozen_split":null,"parent_episo... |
6 | episodes-000000.jsonl | 7dbbb4c5c707afe5a6a70707a13c1e6477c5d79e12ce40b81d507323f8cca679 | simia-agentbench-sft | environment_trajectories | train | content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e | 631 | 125 | unknown | false | [
{
"role": "user",
"text": "You are web shopping.\nI will give you instructions about what to do.\nYou have to follow the instructions.\nEvery round I will give you an observation and a list of available actions, you have to respond an action based on the state and instruction.\nYou can use search action if ... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"7dbbb4c5c707afe5a6a70707a13c1e6477c5d79e12ce40b81d507323f8cca679","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e","frozen_split":null,"parent_episo... |
7 | episodes-000000.jsonl | 37f289e4d5ae248da8e1c9f83243fe2f3fbe5f3443f4fe36e05be8f73f47e29f | simia-agentbench-sft | environment_trajectories | train | content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e | 651 | 130 | unknown | false | [
{
"role": "user",
"text": "You are web shopping.\nI will give you instructions about what to do.\nYou have to follow the instructions.\nEvery round I will give you an observation and a list of available actions, you have to respond an action based on the state and instruction.\nYou can use search action if ... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"37f289e4d5ae248da8e1c9f83243fe2f3fbe5f3443f4fe36e05be8f73f47e29f","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:6f96b29f718c1d9f9ab03d6f154c4e35d68587c3b9b4f512d80f9ababb24c49e","frozen_split":null,"parent_episo... |
8 | episodes-000000.jsonl | 2ee75623c5ff0070a62d3d8b602788aed09f520fd7a4f76f988182cf8d14a729 | simia-agentbench-sft | environment_trajectories | train | content:08abe31c27434aa3fb79415584c48160dcab871f28851dfe517c2115ed8288f8 | 644 | 254 | unknown | false | [
{
"role": "user",
"text": "You are an assistant that will act like a person, I'will play the role of linux(ubuntu) operating system. Your goal is to implement the operations required by me or answer to the question proposed by me. For each of your turn, you should first think what you should do, and then ta... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"2ee75623c5ff0070a62d3d8b602788aed09f520fd7a4f76f988182cf8d14a729","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:08abe31c27434aa3fb79415584c48160dcab871f28851dfe517c2115ed8288f8","frozen_split":null,"parent_episo... |
9 | episodes-000000.jsonl | 7e78ce8433a48fd3972bbf1e112e47f1290dc66ce701639ae7875d9c79ab785f | simia-agentbench-sft | environment_trajectories | train | content:b3baf6079233b5aaa8218ba09c5e30f549b2cb82e55632bd50b4dfc18afbdbbe | 669 | 90 | unknown | false | [
{
"role": "user",
"text": "'''\n<html>\n<div>\n<label id=0 button open navigation drawer>\n<span>Menu</span>\n</label>\n<input id=1 placeholder=\"From\" />\n<input id=2 placeholder=\"To\" />\n<input id=3 type=\"date\" />\n<button id=4 search>Search Flights</button>\n</div>\n</html>\n'''\nBased on the HTML w... | [] | {"actions":[],"control_reference":"","control_seed":"","dependencies":[],"episode_id":"7e78ce8433a48fd3972bbf1e112e47f1290dc66ce701639ae7875d9c79ab785f","lineage":{"ancestor_ids":[],"created_at":null,"family_id":"content:b3baf6079233b5aaa8218ba09c5e30f549b2cb82e55632bd50b4dfc18afbdbbe","frozen_split":null,"parent_episo... |
End of preview. Expand in Data Studio
Megatoken training data
Zstandard-compressed training data: readable Parquet subsets and a compressed project cache restored with the project's download tools. Contact: open a discussion in this repository's Community tab.
Prepared conversations, tool-use and agent trajectories, reasoning traces and code sessions used to train megatoken models. Each row is one episode: a transcript made of role-tagged records, with the tool definitions it used. Episodes were parsed from the public sources listed below and checked. Within each preparation run, episodes are deduplicated across licence groups (the copy from the most permissive source is kept) and split into train, validation and test by family, so related episodes stay on one side. Different runs may contain overlapping episodes and families; consumers combining runs should deduplicate and establish their splits again.
Subsets and licences
The repository's top-level licence is other because the subsets differ. Each subset has one licence group; rows of different groups are never mixed in a file.
| Subset | What it holds | Licence of the subset | Sources | Episodes |
|---|---|---|---|---|
permissive |
Sources under permissive licences (MIT, Apache-2.0, BSD, CC0, CC-BY, ODC-By, ODbL, PDDL, Unlicense or close equivalents). | each source's own permissive licence (see the attribution table) | 1 | 38,178 |
non_commercial |
Sources under non-commercial licences (CC-BY-NC, CC-BY-NC-SA), published under the same non-commercial licence. | CC-BY-NC-4.0 | 1 | 6,677 |
| Compressed cache | Prepared data restored with the project's download tools. | source-specific | not listed | 1 compressed parts, 17.0 MiB |
Licences are the ones the publishers state on each source's Hub page at the revision listed in the tables; we have not verified them independently. Following a source's licence (attribution, share-alike, non-commercial use) is your responsibility when you use its rows. The permissive subset adds no terms of its own; the share_alike subset must be shared under CC-BY-SA-4.0; the non_commercial subset is for non-commercial use only.
Loading
from datasets import load_dataset
rows = load_dataset("gabrielpickard/megatoken-training-data", "permissive", split="train", streaming=True)
| Column | Meaning |
|---|---|
source_id, stream, split, family_id |
the source, the kind of data, the train, validation or test split, and the family that keeps related episodes together |
messages |
readable transcript: a list of {role, text} |
tools |
the tool definitions of the episode (JSON text) |
tokens, scored_positions, outcome, verified_success |
measured when the episode was prepared |
episode_json |
the exact stored episode (payloads in base64) from which megatoken rebuilds its training files |
Attribution
permissive
| Source | Upstream | Revision | Licence | Licence URL | Episodes |
|---|---|---|---|---|---|
| lumos-unified-plan-iterative | ai2lumos/lumos_unified_plan_iterative | e93c0f4856b9 |
apache-2.0 | link | 38,178 |
non_commercial
| Source | Upstream | Revision | Licence | Licence URL | Episodes |
|---|---|---|---|---|---|
| simia-agentbench-sft | Simia-Agent/Simia-AgentBench-SFT-15k | 340f059eda3b |
cc-by-nc-4.0 | link | 6,677 |
Contact and takedown
If you hold rights to material in this repository and want it removed, or if a licence above is wrong: open a discussion in this repository's Community tab. Rows are removed on request.
- Downloads last month
- 118