Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code: DatasetGenerationError
Exception: TypeError
Message: Couldn't cast array of type int64 to null
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1827, in _prepare_split_single
for key, table in generator:
^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
for item in generator(*args, **kwargs):
~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2312, in cast_table_to_schema
cast_array_to_feature(
~~~~~~~~~~~~~~~~~~~~~^
table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
feature,
^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1861, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
~~~~^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2068, in cast_array_to_feature
_c(array.field(name) if name in array_fields else null_array, subfeature)
~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1863, in wrapper
return func(array, *args, **kwargs)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2152, in cast_array_to_feature
return array_cast(
array,
...<2 lines>...
allow_decimal_to_str=allow_decimal_to_str,
)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1863, in wrapper
return func(array, *args, **kwargs)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2014, in array_cast
raise TypeError(f"Couldn't cast array of type {_short_str(array.type)} to {_short_str(pa_type)}")
TypeError: Couldn't cast array of type int64 to null
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1880, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
answers list | canonical_query string | competing_paths list | composition list | conflicting_paths list | difficulty int64 | example_id string | generator_version string | graph dict | intent_id string | negative_answers list | query dict | seed int64 | stage_name string | status string | supporting_paths list | task_type string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
[
"e2872753_31"
] | find_entities|Which company employs e5342235_1?|e5342235_1||||||||1|2|None|entity| | [
{
"score": 0,
"steps": [
{
"edge_attributes": {
"weight": "2"
},
"relation_id": "reports_to",
"source_id": "e5342235_1",
"source_type": "Person",
"target_id": "e5277981_3",
"target_type": "Person"
}
],
"verified": false
... | [
"works_at"
] | [] | 0 | graph-synthetic-3-1-0 | graph-synthetic-3 | {
"edges": [
{
"attributes": {
"weight": "1"
},
"relation_id": "works_at",
"relation_name": "works_at",
"source_id": "e5342235_1",
"source_type": "Person",
"target_id": "e2872753_31",
"target_type": "Company"
},
{
"attributes": {
"weigh... | employer | [
"e5277981_3",
"e0109187_40",
"e6746097_37",
"e2418262_36"
] | {
"allowed_relations": [],
"answer_mode": "entity",
"exact_hops": null,
"forbidden_relations": [],
"forbidden_types": [],
"max_hops": 2,
"min_hops": 1,
"operation": "find_entities",
"required_intermediate_types": [],
"required_relation_sequence": [],
"required_relations": [],
"required_types": [... | 1 | graph_encoding | supported | [
{
"score": 0,
"steps": [
{
"edge_attributes": {
"weight": "1"
},
"relation_id": "works_at",
"source_id": "e5342235_1",
"source_type": "Person",
"target_id": "e2872753_31",
"target_type": "Company"
}
],
"verified": false
... | find_entities |
[
"e9217203_1"
] | find_entities|Name the supervisor of e0108567_65.|e0108567_65||||||||1|2|None|entity| | [
{
"score": 0,
"steps": [
{
"edge_attributes": {
"weight": "0"
},
"relation_id": "member_of",
"source_id": "e0108567_65",
"source_type": "Person",
"target_id": "e8250271_75",
"target_type": "Team"
}
],
"verified": false
}... | [
"reports_to"
] | [] | 0 | graph-synthetic-3-2-0 | graph-synthetic-3 | {
"edges": [
{
"attributes": {
"weight": "3"
},
"relation_id": "works_at",
"relation_name": "works_at",
"source_id": "e7303929_13",
"source_type": "Person",
"target_id": "e8270070_2",
"target_type": "Company"
},
{
"attributes": {
"weigh... | manager | [
"e8250271_75",
"e3559657_30",
"e6906336_42",
"e5587071_84"
] | {
"allowed_relations": [],
"answer_mode": "entity",
"exact_hops": null,
"forbidden_relations": [],
"forbidden_types": [],
"max_hops": 2,
"min_hops": 1,
"operation": "find_entities",
"required_intermediate_types": [],
"required_relation_sequence": [],
"required_relations": [],
"required_types": [... | 2 | graph_encoding | supported | [
{
"score": 0,
"steps": [
{
"edge_attributes": {
"weight": "0"
},
"relation_id": "reports_to",
"source_id": "e0108567_65",
"source_type": "Person",
"target_id": "e9217203_1",
"target_type": "Person"
}
],
"verified": false
... | find_entities |
[
"e0262922_61"
] | find_entities|Name the employer of e8929016_48.|e8929016_48||||||||1|2|None|entity| | [
{
"score": 0,
"steps": [
{
"edge_attributes": {
"weight": "2"
},
"relation_id": "reports_to",
"source_id": "e8929016_48",
"source_type": "Person",
"target_id": "e3303122_50",
"target_type": "Person"
}
],
"verified": false
... | [
"works_at"
] | [] | 0 | graph-synthetic-3-3-0 | graph-synthetic-3 | {
"edges": [
{
"attributes": {
"weight": "2"
},
"relation_id": "works_at",
"relation_name": "works_at",
"source_id": "e6308367_0",
"source_type": "Person",
"target_id": "e2839754_5",
"target_type": "Company"
},
{
"attributes": {
"weight... | employer | [
"e3303122_50"
] | {
"allowed_relations": [],
"answer_mode": "entity",
"exact_hops": null,
"forbidden_relations": [],
"forbidden_types": [],
"max_hops": 2,
"min_hops": 1,
"operation": "find_entities",
"required_intermediate_types": [],
"required_relation_sequence": [],
"required_relations": [],
"required_types": [... | 3 | graph_encoding | supported | [
{
"score": 0,
"steps": [
{
"edge_attributes": {
"weight": "2"
},
"relation_id": "works_at",
"source_id": "e8929016_48",
"source_type": "Person",
"target_id": "e0262922_61",
"target_type": "Company"
}
],
"verified": false
... | find_entities |
[
"e9371458_3"
] | find_entities|Name the supervisor of e7901302_47.|e7901302_47||||||||1|2|None|entity| | [
{
"score": 0,
"steps": [
{
"edge_attributes": {
"weight": "4"
},
"relation_id": "works_at",
"source_id": "e7901302_47",
"source_type": "Person",
"target_id": "e3905236_20",
"target_type": "Company"
}
],
"verified": false
... | [
"reports_to"
] | [] | 0 | graph-synthetic-3-4-0 | graph-synthetic-3 | {
"edges": [
{
"attributes": {
"weight": "1"
},
"relation_id": "works_at",
"relation_name": "works_at",
"source_id": "e9371458_3",
"source_type": "Person",
"target_id": "e8073850_10",
"target_type": "Company"
},
{
"attributes": {
"weigh... | manager | [
"e3905236_20",
"e5083432_54",
"e1639461_6",
"e1332402_5",
"e6455471_49"
] | {
"allowed_relations": [],
"answer_mode": "entity",
"exact_hops": null,
"forbidden_relations": [],
"forbidden_types": [],
"max_hops": 2,
"min_hops": 1,
"operation": "find_entities",
"required_intermediate_types": [],
"required_relation_sequence": [],
"required_relations": [],
"required_types": [... | 4 | graph_encoding | supported | [
{
"score": 0,
"steps": [
{
"edge_attributes": {
"weight": "3"
},
"relation_id": "reports_to",
"source_id": "e7901302_47",
"source_type": "Person",
"target_id": "e9371458_3",
"target_type": "Person"
}
],
"verified": false
... | find_entities |
[
"e1764667_25"
] | find_entities|Who does e2308002_75 report to?|e2308002_75||||||||1|2|None|entity| | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"1"},"relation_id":"member_of","source_id":"e230(...TRUNCATED) | [
"reports_to"
] | [] | 0 | graph-synthetic-3-5-0 | graph-synthetic-3 | {"edges":[{"attributes":{"weight":"2"},"relation_id":"works_at","relation_name":"works_at","source_i(...TRUNCATED) | manager | ["e9887315_61","e3766793_48","e5771888_24","e5033514_17","e9478906_88","e9113879_59","e5950604_58","(...TRUNCATED) | {"allowed_relations":[],"answer_mode":"entity","exact_hops":null,"forbidden_relations":[],"forbidden(...TRUNCATED) | 5 | graph_encoding | supported | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"0"},"relation_id":"reports_to","source_id":"e23(...TRUNCATED) | find_entities |
[
"e5224132_3"
] | find_entities|Name the supervisor of e8890321_78.|e8890321_78||||||||1|2|None|entity| | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"3"},"relation_id":"manages","source_id":"e88903(...TRUNCATED) | [
"reports_to"
] | [] | 0 | graph-synthetic-3-6-0 | graph-synthetic-3 | {"edges":[{"attributes":{"weight":"0"},"relation_id":"works_at","relation_name":"works_at","source_i(...TRUNCATED) | manager | [
"e2281940_32",
"e7235507_2",
"e6766997_23",
"e6873910_59",
"e5199139_92",
"e4372377_33"
] | {"allowed_relations":[],"answer_mode":"entity","exact_hops":null,"forbidden_relations":[],"forbidden(...TRUNCATED) | 6 | graph_encoding | supported | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"3"},"relation_id":"reports_to","source_id":"e88(...TRUNCATED) | find_entities |
[
"e5271263_8"
] | find_entities|Which company employs e3453827_11?|e3453827_11||||||||1|2|None|entity| | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"1"},"relation_id":"reports_to","source_id":"e34(...TRUNCATED) | [
"works_at"
] | [] | 0 | graph-synthetic-3-7-0 | graph-synthetic-3 | {"edges":[{"attributes":{"weight":"0"},"relation_id":"works_at","relation_name":"works_at","source_i(...TRUNCATED) | employer | [
"e6093875_9",
"e3018865_44",
"e3432513_60",
"e8014632_21",
"e4163613_53",
"e5787000_26"
] | {"allowed_relations":[],"answer_mode":"entity","exact_hops":null,"forbidden_relations":[],"forbidden(...TRUNCATED) | 7 | graph_encoding | supported | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"3"},"relation_id":"works_at","source_id":"e3453(...TRUNCATED) | find_entities |
[
"e7665918_11"
] | find_entities|Which company employs e3558944_3?|e3558944_3||||||||1|2|None|entity| | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"1"},"relation_id":"member_of","source_id":"e355(...TRUNCATED) | [
"works_at"
] | [] | 0 | graph-synthetic-3-8-0 | graph-synthetic-3 | {"edges":[{"attributes":{"weight":"2"},"relation_id":"works_at","relation_name":"works_at","source_i(...TRUNCATED) | employer | [
"e5084856_23",
"e3918187_7",
"e9989872_63",
"e5489918_5",
"e4625441_15",
"e7783466_19"
] | {"allowed_relations":[],"answer_mode":"entity","exact_hops":null,"forbidden_relations":[],"forbidden(...TRUNCATED) | 8 | graph_encoding | supported | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"2"},"relation_id":"works_at","source_id":"e3558(...TRUNCATED) | find_entities |
[
"e9491263_71"
] | find_entities|Which person is the manager of e7447717_46?|e7447717_46||||||||1|2|None|entity| | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"4"},"relation_id":"works_at","source_id":"e7447(...TRUNCATED) | [
"reports_to"
] | [] | 0 | graph-synthetic-3-9-0 | graph-synthetic-3 | {"edges":[{"attributes":{"weight":"1"},"relation_id":"works_at","relation_name":"works_at","source_i(...TRUNCATED) | manager | [
"e6641327_48",
"e3266417_75",
"e7925791_56",
"e1207268_74",
"e0573550_47"
] | {"allowed_relations":[],"answer_mode":"entity","exact_hops":null,"forbidden_relations":[],"forbidden(...TRUNCATED) | 9 | graph_encoding | supported | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"0"},"relation_id":"reports_to","source_id":"e74(...TRUNCATED) | find_entities |
[
"e9202620_22"
] | find_entities|Name the employer of e7866538_35.|e7866538_35||||||||1|2|None|entity| | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"3"},"relation_id":"reports_to","source_id":"e78(...TRUNCATED) | [
"works_at"
] | [] | 0 | graph-synthetic-3-10-0 | graph-synthetic-3 | {"edges":[{"attributes":{"weight":"3"},"relation_id":"works_at","relation_name":"works_at","source_i(...TRUNCATED) | employer | [
"e1127181_6"
] | {"allowed_relations":[],"answer_mode":"entity","exact_hops":null,"forbidden_relations":[],"forbidden(...TRUNCATED) | 10 | graph_encoding | supported | [{"score":0.0,"steps":[{"edge_attributes":{"weight":"2"},"relation_id":"works_at","source_id":"e7866(...TRUNCATED) | find_entities |
End of preview.
graph-reasoner1-trainning-data
Synthetic multi-hop graph-reasoning examples: the corpus frontal-labs/graph-reasoner-1 was trained and evaluated on. Each row is a self-contained typed graph, a query, the gold answers, gold evidence paths, and same-length competing paths.
39,000 examples, 131 MB gzipped.
| split | examples | what it is |
|---|---|---|
train |
30,000 | the curriculum stream training consumes, all 15 stages |
validation |
3,000 | stage-balanced, graph-disjoint from train |
test |
3,000 | stage-balanced, graph-disjoint from train and validation |
holdout_unseen_topology |
600 | graph families not seen in training |
holdout_unseen_entities |
600 | entirely unseen entities |
holdout_unseen_composition |
600 | relation compositions withheld from training |
holdout_unseen_ontology |
600 | relation labels renamed to opaque tokens |
holdout_unseen_path_distribution |
600 | dense, hub-heavy graphs |
Splits are graph-disjoint by construction: each draws from a reserved seed range
(train=1, validation=4_000_000, test=8_000_000, holdouts 12_000_000).
A sample of a generator, not a fixed corpus
This is the honest framing. The real artifact is the generator in the source repo; this is a reproducible sample of it. Regenerate this exact data, or any other size, with:
python scripts/generate_synthetic.py --output data.jsonl --count 30000 --seed 1
generator_version is recorded in counts.json and on every row. PLAN §42 called for a
10M-example corpus; that was never materialised, and this 39k sample is not it.
Each row
| field | meaning |
|---|---|
graph |
nodes with types, edges with relation labels, attributes |
canonical_query / intent_id |
the query, and which of 15 intents produced it |
answers |
gold answer entity ids |
supporting_paths |
gold evidence paths |
competing_paths |
same-length wrong routes (the hard negatives) |
composition |
the gold relation chain — see the warning below |
status |
supported, insufficient_evidence, contradicted, unsupported |
difficulty, generator_version, example_id |
provenance |
Do not feed composition to the model at inference
It is the answer key. The queries are deliberately well-posed: they state what is
wanted without revealing the relation chain, its length, or the answer type. Supplying
composition reduces the task to following a given path, at which a breadth-first
traversal scores a perfect 1.000. Earlier iterations of this benchmark leaked it and
saturated at 1.000 exact match; three CI guards now prevent that regression.
The ontology
10 entity types, 18 relations with type signatures and a functional flag
(distinguishing "has exactly one value" from genuinely multi-valued, which is what
separates a contradiction from a multi-valued answer), and 15 query intents with
paraphrase variants.
Known limitation of the benchmark
holdout_unseen_ontology renames relations to opaque tokens (R16961) while keeping type
names canonical, so a relation's role must be inferred from its type signature. It is
well-posed but unsolved: the published model scores 0.158 exact match with PAT
discrimination at 0.577 against a 0.5 floor. Treat it as an open problem, not a
calibration target.
Apache-2.0. Generated by graph-reasoning-model; no human or scraped data.
- Downloads last month
- 36