This dataset is used to train the bekko-system-one-v0 model family.
Public Hugging Face dataset distribution.
from datasets import load_dataset
dataset = load_dataset("hotchpotch/bekko-system-one-dataset-v0", "laya__typed_decisions")
test = dataset["test"]
Select a configuration explicitly. Parquet uses Zstandard compression and dictionary encoding; all decoded values and row order match the local Arrow release. Only training-manifest.json defines active membership. Configurations listed in quarantine-manifest.json are retained for audit and must not enter evaluation. Source attribution and license scope are retained in sources.json, SOURCES.md, dataset_metadata.json and _sources/. README has one leading YAML block listing every config and split. Hub manifests point to Parquet files; _sources/local-arrow/ preserves the historical Arrow manifests.
Typed decisions over shared state, with Choice, Noul and numeric Score supervision.
Train uses a 300,000-judgment cap per subset with complete cases/groups. Validation is sampled from source validation. Test matches the active S1MB dataset. Use training-manifest.json for membership.
| Role | Subsets | Cases | Judgments |
|---|---|---|---|
| train | 153 | 6,589,188 | 8,421,771 |
| evaluation | 125 | 22,210 | 41,938 |
| unlabeled | 0 | 0 | 0 |
| Column | Purpose |
|---|---|
| input | Shared state and typed decisions; the only inference input. |
| targets | Aligned target distributions and annotation metadata. |
| split | train, validation or test. |
| case_id | Stable case identifier. |
| group_id | Related-case grouping. |
| language | Language code. |
| input_hash | Stable input identity for audits. |
Score values come from the explicit numeric criteria, not candidate indices. Noul retains authored false/true definitions. Preserve soft targets and candidate order. Repeated strings use physical dictionary encoding; decoded values remain strings.
See sources.json, SOURCES.md and dataset_metadata.json for upstream datasets, revisions, attribution and license scope. The laya__ namespace denotes the Laya-source conversion pipeline; it does not imply Laya authored those upstream datasets. laya__typed_decisions contains four synthetic workflows published by LocalLLaMA. No blanket relicensing is applied. namespace-map.json records names; identifiers remain stable.
Counts below are cases in this release, not upstream dataset sizes or judgment counts. A case can contain multiple judgments. Counts use the active training-manifest.json; 0 means that split is not included. MNLI validation combines validation_matched and validation_mismatched. Source links include acquisition datasets and original dataset references where recorded; they do not imply direct acquisition from each original.
Licenses below describe the original datasets. Follow the links for their full terms. Unknown means the data license has not been confirmed.
Each subset follows its original dataset's license and usage terms. Check the linked original sources before using or redistributing the data. This collection does not change those licenses.
Where noted, code, annotations and source documents have different terms. Detailed source records are in sources.json. An unknown license is not permission to reuse the data.
For Open-Jev, retain its acquisition attribution and third-party notices. In open_jev__customer-control-v1, the generated conversations have a CC0-1.0 declaration, but the upstream question descriptions have no verified license and are not relicensed as CC0. Authored S1MB generalization subsets retain their recorded not separately declared status. See sources.json, SOURCES.md and the retained upstream notices for scope and qualifications.
Removed two complete training cases whose full inputs match active S1MB test cases after normalizing whitespace: contract_nli:default:train:366 (17 decisions) and enron_spam:default:train:18716 (1 decision, configuration laya__enron_spam). ContractNLI also has an upstream train/test target disagreement on nda-1. Test and validation membership, retained rows, candidate order and targets are unchanged. See decontamination-20261008.json for the parent revision, verification and removed IDs. This is a targeted deletion, not a claim of comprehensive semantic decontamination. Historical Arrow/source/export statistics remain in _sources/; current counts are in the active manifests and release_splits.
This dataset is used to train the bekko-system-one-v0 model family.
Public Hugging Face dataset distribution.
from datasets import load_dataset
dataset = load_dataset("hotchpotch/bekko-system-one-dataset-v0", "laya__typed_decisions")
test = dataset["test"]
Select a configuration explicitly. Parquet uses Zstandard compression and dictionary encoding; all decoded values and row order match the local Arrow release. Only training-manifest.json defines active membership. Configurations listed in quarantine-manifest.json are retained for audit and must not enter evaluation. Source attribution and license scope are retained in sources.json, SOURCES.md, dataset_metadata.json and _sources/. README has one leading YAML block listing every config and split. Hub manifests point to Parquet files; _sources/local-arrow/ preserves the historical Arrow manifests.
Typed decisions over shared state, with Choice, Noul and numeric Score supervision.
Train uses a 300,000-judgment cap per subset with complete cases/groups. Validation is sampled from source validation. Test matches the active S1MB dataset. Use training-manifest.json for membership.
| Role | Subsets | Cases | Judgments |
|---|---|---|---|
| train | 153 | 6,589,188 | 8,421,771 |
| evaluation | 125 | 22,210 | 41,938 |
| unlabeled | 0 | 0 | 0 |
| Column | Purpose |
|---|---|
| input | Shared state and typed decisions; the only inference input. |
| targets | Aligned target distributions and annotation metadata. |
| split | train, validation or test. |
| case_id | Stable case identifier. |
| group_id | Related-case grouping. |
| language | Language code. |
| input_hash | Stable input identity for audits. |
Score values come from the explicit numeric criteria, not candidate indices. Noul retains authored false/true definitions. Preserve soft targets and candidate order. Repeated strings use physical dictionary encoding; decoded values remain strings.
See sources.json, SOURCES.md and dataset_metadata.json for upstream datasets, revisions, attribution and license scope. The laya__ namespace denotes the Laya-source conversion pipeline; it does not imply Laya authored those upstream datasets. laya__typed_decisions contains four synthetic workflows published by LocalLLaMA. No blanket relicensing is applied. namespace-map.json records names; identifiers remain stable.
Counts below are cases in this release, not upstream dataset sizes or judgment counts. A case can contain multiple judgments. Counts use the active training-manifest.json; 0 means that split is not included. MNLI validation combines validation_matched and validation_mismatched. Source links include acquisition datasets and original dataset references where recorded; they do not imply direct acquisition from each original.
Licenses below describe the original datasets. Follow the links for their full terms. Unknown means the data license has not been confirmed.
Each subset follows its original dataset's license and usage terms. Check the linked original sources before using or redistributing the data. This collection does not change those licenses.
Where noted, code, annotations and source documents have different terms. Detailed source records are in sources.json. An unknown license is not permission to reuse the data.
For Open-Jev, retain its acquisition attribution and third-party notices. In open_jev__customer-control-v1, the generated conversations have a CC0-1.0 declaration, but the upstream question descriptions have no verified license and are not relicensed as CC0. Authored S1MB generalization subsets retain their recorded not separately declared status. See sources.json, SOURCES.md and the retained upstream notices for scope and qualifications.
Removed two complete training cases whose full inputs match active S1MB test cases after normalizing whitespace: contract_nli:default:train:366 (17 decisions) and enron_spam:default:train:18716 (1 decision, configuration laya__enron_spam). ContractNLI also has an upstream train/test target disagreement on nda-1. Test and validation membership, retained rows, candidate order and targets are unchanged. See decontamination-20261008.json for the parent revision, verification and removed IDs. This is a targeted deletion, not a claim of comprehensive semantic decontamination. Historical Arrow/source/export statistics remain in _sources/; current counts are in the active manifests and release_splits.