Dataset Preview
Duplicate
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:    CastError
Message:      Couldn't cast
action: struct<execution_graph: struct<edges: list<item: struct<from: string, kind: string, to: string>>, no (... 121 chars omitted)
  child 0, execution_graph: struct<edges: list<item: struct<from: string, kind: string, to: string>>, nodes: list<item: string>>
      child 0, edges: list<item: struct<from: string, kind: string, to: string>>
          child 0, item: struct<from: string, kind: string, to: string>
              child 0, from: string
              child 1, kind: string
              child 2, to: string
      child 1, nodes: list<item: string>
          child 0, item: string
  child 1, kind: string
  child 2, message: string
  child 3, tool_calls: list<item: struct<arguments: string, tool: string>>
      child 0, item: struct<arguments: string, tool: string>
          child 0, arguments: string
          child 1, tool: string
grounding: string
history: list<item: struct<ask: string, reply: string, tools: list<item: null>>>
  child 0, item: struct<ask: string, reply: string, tools: list<item: null>>
      child 0, ask: string
      child 1, reply: string
      child 2, tools: list<item: null>
          child 0, item: null
negative_controls: list<item: string>
  child 0, item: string
reference: string
task: struct<description: string, id: string, initial_state: string, required: string, tool_names: list<it (... 12 chars omitted)
  child 0, description: string
  child 1, id: string
  child 2, initial_state: string
  child 3, required: string
  child 4, tool_names: list<item: string>
      child 0, item: string
to
{'action': {'execution_graph': {'edges': List({'from': Value('string'), 'kind': Value('string'), 'to': Value('string')}), 'nodes': List(Json(decode=True))}, 'kind': Value('string'), 'message': Value('string'), 'tool_calls': List({'arguments': Json(decode=True), 'tool': Value('string')})}, 'negative_controls': List(Json(decode=True)), 'reference': Json(decode=True), 'task': {'description': Value('string'), 'id': Value('string'), 'initial_state': Json(decode=True), 'required': Json(decode=True), 'tool_names': List(Value('string'))}, 'grounding': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1835, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 621, in wrapped
                  for item in generator(*args, **kwargs):
                              ~~~~~~~~~^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, 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 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              action: struct<execution_graph: struct<edges: list<item: struct<from: string, kind: string, to: string>>, no (... 121 chars omitted)
                child 0, execution_graph: struct<edges: list<item: struct<from: string, kind: string, to: string>>, nodes: list<item: string>>
                    child 0, edges: list<item: struct<from: string, kind: string, to: string>>
                        child 0, item: struct<from: string, kind: string, to: string>
                            child 0, from: string
                            child 1, kind: string
                            child 2, to: string
                    child 1, nodes: list<item: string>
                        child 0, item: string
                child 1, kind: string
                child 2, message: string
                child 3, tool_calls: list<item: struct<arguments: string, tool: string>>
                    child 0, item: struct<arguments: string, tool: string>
                        child 0, arguments: string
                        child 1, tool: string
              grounding: string
              history: list<item: struct<ask: string, reply: string, tools: list<item: null>>>
                child 0, item: struct<ask: string, reply: string, tools: list<item: null>>
                    child 0, ask: string
                    child 1, reply: string
                    child 2, tools: list<item: null>
                        child 0, item: null
              negative_controls: list<item: string>
                child 0, item: string
              reference: string
              task: struct<description: string, id: string, initial_state: string, required: string, tool_names: list<it (... 12 chars omitted)
                child 0, description: string
                child 1, id: string
                child 2, initial_state: string
                child 3, required: string
                child 4, tool_names: list<item: string>
                    child 0, item: string
              to
              {'action': {'execution_graph': {'edges': List({'from': Value('string'), 'kind': Value('string'), 'to': Value('string')}), 'nodes': List(Json(decode=True))}, 'kind': Value('string'), 'message': Value('string'), 'tool_calls': List({'arguments': Json(decode=True), 'tool': Value('string')})}, 'negative_controls': List(Json(decode=True)), 'reference': Json(decode=True), 'task': {'description': Value('string'), 'id': Value('string'), 'initial_state': Json(decode=True), 'required': Json(decode=True), 'tool_names': List(Value('string'))}, 'grounding': Value('string')}
              because column names don't match
              
              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 1379, 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 956, 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 1702, 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 1887, in _prepare_split_single
                  raise DatasetGenerationError("An error occurred while generating the dataset") from e
              datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset

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.

action
dict
negative_controls
list
reference
unknown
task
dict
grounding
string
{ "execution_graph": { "edges": [ { "from": "881cb92a-1d6a-4a4d-a3fb-f9dd8e2842ae", "kind": "data_flow", "to": "04c2ce73-290b-407a-a64e-f2d35b9a9bc3" } ], "nodes": [ { "id": "881cb92a-1d6a-4a4d-a3fb-f9dd8e2842ae", "kind": "connect", "name":...
[ { "calibrated": true, "calibration_loop": true, "connect": true, "explain": "none", "leakage_free": true, "model_kind": "llm" }, { "calibrated": true, "calibration_loop": true, "connect": true, "explain": "shap", "leakage_free": false, "model_kind": "gbdt" }, ...
{ "calibrated": true, "calibration_loop": true, "connect": true, "explain": "shap", "leakage_free": true, "model_kind": "gbdt" }
{ "description": "Build a churn-risk pipeline over the account events + communications: connect the source, engineer leakage-free point-in-time features, train a gradient-boosted model with SHAP explanations, and serve calibrated risk scores with a score-vs-outcome calibration loop.", "id": "uc1-churn", "initial_...
null
{ "execution_graph": { "edges": [ { "from": "15982231-3fca-4bbb-8254-3264f7ad328d", "kind": "data_flow", "to": "2e4b24f0-ba95-4f48-92d5-733eaa58ec60" }, { "from": "2e4b24f0-ba95-4f48-92d5-733eaa58ec60", "kind": "data_flow", "to": "ec3a009e-6e31-4c4...
[ { "embeddings": false, "grounded_in_data": true, "index": false, "ranked": true, "retrieval": false }, { "embeddings": false, "grounded_in_data": true, "index": true, "index_kind": "bm25f", "ranked": false, "retrieval": true }, { "embeddings": false, "grou...
{ "embeddings": false, "grounded_in_data": true, "index": true, "index_kind": "bm25f", "ranked": true, "retrieval": true }
{ "description": "Make the documents searchable: connect the source, build a full-text (BM25F) search index over them, and serve ranked results grounded in the indexed data.", "id": "uc-search", "initial_state": { "data.documents": true }, "required": { "grounded_in_data": true, "index": true, ...
null
{ "execution_graph": { "edges": [ { "from": "56c01d9a-18a6-49bf-bea0-53c532aedd80", "kind": "data_flow", "to": "6d910ddb-77e0-4fd3-9a09-2af683a13d91" }, { "from": "6d910ddb-77e0-4fd3-9a09-2af683a13d91", "kind": "data_flow", "to": "655d71b2-703c-4c5...
[ { "calibrated": false, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }, { "calibrated": true, "clean": false, "connect": true, ...
{ "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Score our leads + opportunities so sales works the ones most likely to convert.", "id": "S-2", "initial_state": {}, "required": { "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "p...
null
{ "execution_graph": { "edges": [ { "from": "d3c96e09-1ae4-4783-9235-47bc37b58fb6", "kind": "data_flow", "to": "df98ef63-ea79-4ea2-8c32-0a5d84acd2f1" }, { "from": "df98ef63-ea79-4ea2-8c32-0a5d84acd2f1", "kind": "data_flow", "to": "de8982a1-b83f-487...
[ { "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "regression", "point_in_time": true, "serve": true, "temporal_validation": true }, { "calibrated": false, "clean": false, "connect": tru...
{ "calibrated": false, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "regression", "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Predict each customer's lifetime value from their history.", "id": "S-3", "initial_state": {}, "required": { "calibrated": false, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "regression", "point_in_time":...
null
{ "execution_graph": { "edges": [ { "from": "42f3c104-15a2-4a16-82d5-5a3b5bee4fe5", "kind": "data_flow", "to": "a2b126e0-8da7-4d1d-a9ac-d1ecd21c164c" }, { "from": "a2b126e0-8da7-4d1d-a9ac-d1ecd21c164c", "kind": "data_flow", "to": "57aa1347-0cb9-436...
[ { "calibrated": false, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }, { "calibrated": true, "clean": false, "connect": true, ...
{ "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Prioritize which overdue accounts to work first for collections.", "id": "S-4", "initial_state": {}, "required": { "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": ...
null
{ "execution_graph": { "edges": [ { "from": "ae59f3bf-2b8b-48c9-8787-0c851bacdaf5", "kind": "data_flow", "to": "e99c9c53-3f81-4ae3-8400-e0f7ec2e5981" }, { "from": "e99c9c53-3f81-4ae3-8400-e0f7ec2e5981", "kind": "data_flow", "to": "70e1d647-d5d0-433...
[ { "act": false, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "survival", "point_in_time": true, "serve": true, "temporal_validation": true }, { "act": true, ...
{ "act": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "survival", "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Predict which machines are likely to fail so we can service them first.", "id": "S-8", "initial_state": {}, "required": { "act": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": t...
null
{ "execution_graph": { "edges": [ { "from": "e0c912c4-9472-46b4-955c-0fc123306eb7", "kind": "data_flow", "to": "89ea7d92-c583-4321-a007-9dd9447d8c6e" }, { "from": "89ea7d92-c583-4321-a007-9dd9447d8c6e", "kind": "data_flow", "to": "ce031d0e-d47c-4ed...
[ { "act": false, "anomaly": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }, { ...
{ "act": true, "anomaly": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Flag likely fraud on transactions for review, with reasons, without over-blocking.", "id": "S-10", "initial_state": {}, "required": { "act": true, "anomaly": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "le...
null
{ "execution_graph": { "edges": [ { "from": "c2d4293b-3c70-4d5c-b824-5ef44056b7a4", "kind": "data_flow", "to": "ddd6c76c-290b-4950-b57c-ea318d7a6a0d" }, { "from": "ddd6c76c-290b-4950-b57c-ea318d7a6a0d", "kind": "data_flow", "to": "a2e1f664-3b7d-4a1...
[ { "act": false, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }, { "act": true, "a...
{ "act": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Triage incoming claims by risk + route the risky ones for review.", "id": "S-12", "initial_state": {}, "required": { "act": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, ...
null
{ "execution_graph": { "edges": [ { "from": "776688c5-16fc-4b97-b49a-d213cb45bb4a", "kind": "data_flow", "to": "53d63202-b2ce-4006-b4d8-b27e325fdd8e" }, { "from": "53d63202-b2ce-4006-b4d8-b27e325fdd8e", "kind": "data_flow", "to": "c8750f30-8641-482...
[ { "backtest": false, "clean": true, "connect": true, "leakage_free": true, "model": true, "model_kind": "forecast", "serve": true, "temporal_validation": true }, { "backtest": true, "clean": false, "connect": true, "leakage_free": true, "model": true, "mod...
{ "backtest": true, "clean": true, "connect": true, "leakage_free": true, "model": true, "model_kind": "forecast", "serve": true, "temporal_validation": true }
{ "description": "Forecast demand per product/location for the coming weeks.", "id": "S-5", "initial_state": {}, "required": { "backtest": true, "clean": true, "connect": true, "leakage_free": true, "model": true, "model_kind": "forecast", "serve": true, "temporal_validation": tr...
null
{ "execution_graph": { "edges": [ { "from": "b9bcb2aa-2099-4f20-8077-cac11f105a87", "kind": "data_flow", "to": "3fe82942-d6a8-4b0b-9c69-8dff79153c69" }, { "from": "3fe82942-d6a8-4b0b-9c69-8dff79153c69", "kind": "data_flow", "to": "364584c8-91a7-454...
[ { "act": false, "approval_gated": true, "connect": true, "model": true, "model_kind": "forecast", "optimization": true, "serve": true, "temporal_validation": true }, { "act": true, "approval_gated": false, "connect": true, "model": true, "model_kind": "forecas...
{ "act": true, "approval_gated": true, "connect": true, "model": true, "model_kind": "forecast", "optimization": true, "serve": true, "temporal_validation": true }
{ "description": "Turn the demand forecast into replenishment orders under our stock constraints.", "id": "S-6", "initial_state": {}, "required": { "act": true, "approval_gated": true, "connect": true, "model": true, "model_kind": "forecast", "optimization": true, "serve": true, ...
null
{ "execution_graph": { "edges": [ { "from": "68752085-e220-4fe2-b7a2-4be2ed442582", "kind": "data_flow", "to": "8c5ec850-f0ef-47c6-8385-6dcce959f651" }, { "from": "8c5ec850-f0ef-47c6-8385-6dcce959f651", "kind": "data_flow", "to": "3365f747-d2b4-469...
[ { "clean": false, "connect": true, "drivers": true, "explain": "shap", "model": true, "model_kind": "anomaly", "serve": true }, { "clean": true, "connect": false, "drivers": true, "explain": "shap", "model": true, "model_kind": "anomaly", "serve": true }...
{ "clean": true, "connect": true, "drivers": true, "explain": "shap", "model": true, "model_kind": "anomaly", "serve": true }
{ "description": "Tell us when a KPI moves abnormally and what's driving it.", "id": "S-7", "initial_state": {}, "required": { "clean": true, "connect": true, "drivers": true, "explain": "shap", "model": true, "model_kind": "anomaly", "serve": true }, "tool_names": [ "connect...
null
{ "execution_graph": { "edges": [ { "from": "86f64bd5-50b6-4476-b37f-c95e1a2717ec", "kind": "data_flow", "to": "210fd409-cfc1-4b78-9f16-e83e30689a2d" }, { "from": "210fd409-cfc1-4b78-9f16-e83e30689a2d", "kind": "data_flow", "to": "11ddc128-e001-45d...
[ { "act": false, "approval_gated": true, "clean": true, "connect": true, "holdout": true, "incrementality": true, "leakage_free": true, "model": true, "model_kind": "uplift", "serve": true }, { "act": true, "approval_gated": false, "clean": true, "connect":...
{ "act": true, "approval_gated": true, "clean": true, "connect": true, "holdout": true, "incrementality": true, "leakage_free": true, "model": true, "model_kind": "uplift", "serve": true }
{ "description": "Recommend the next best action per customer, and only act where it changes the outcome.", "id": "S-11", "initial_state": {}, "required": { "act": true, "approval_gated": true, "clean": true, "connect": true, "holdout": true, "incrementality": true, "leakage_free": t...
null
{ "execution_graph": { "edges": [ { "from": "9e92f7a5-1d5b-4b00-af61-e52196cf89a3", "kind": "data_flow", "to": "ce24fd5a-8c2b-4a6f-b696-d686f93be775" }, { "from": "ce24fd5a-8c2b-4a6f-b696-d686f93be775", "kind": "data_flow", "to": "d1b4a105-2739-46e...
[ { "connect": false, "embeddings": true, "grounded_in_data": true, "index": true, "index_kind": "semantic", "ranked": true, "retrieval": true, "serve": true }, { "connect": true, "embeddings": false, "grounded_in_data": true, "index": true, "index_kind": "seman...
{ "connect": true, "embeddings": true, "grounded_in_data": true, "index": true, "index_kind": "semantic", "ranked": true, "retrieval": true, "serve": true }
{ "description": "Answer questions over our documents with cited passages, grounded in our data.", "id": "M-10", "initial_state": {}, "required": { "connect": true, "embeddings": true, "grounded_in_data": true, "index": true, "index_kind": "semantic", "ranked": true, "retrieval": tru...
null
{ "execution_graph": { "edges": [ { "from": "944ab36c-d021-4ac6-8352-bc9d582a75ac", "kind": "data_flow", "to": "a224258e-a435-428e-a2ce-95ac9b2d3e2a" }, { "from": "a224258e-a435-428e-a2ce-95ac9b2d3e2a", "kind": "data_flow", "to": "22563d0a-0ad6-418...
[ { "clean": false, "connect": true, "extraction": true, "grounded_in_data": true, "index": true, "index_kind": "semantic", "model": true, "model_kind": "document_intelligence", "serve": true }, { "clean": true, "connect": false, "extraction": true, "grounded_in...
{ "clean": true, "connect": true, "extraction": true, "grounded_in_data": true, "index": true, "index_kind": "semantic", "model": true, "model_kind": "document_intelligence", "serve": true }
{ "description": "Extract the key fields from our invoices/contracts (touchless AP).", "id": "M-11", "initial_state": {}, "required": { "clean": true, "connect": true, "extraction": true, "grounded_in_data": true, "index": true, "index_kind": "semantic", "model": true, "model_kin...
null
{ "execution_graph": { "edges": [ { "from": "bacc2642-a317-41fd-b11a-68f1e1d7a967", "kind": "data_flow", "to": "2a18f49f-67d1-416b-b75d-181b7fbda1d6" }, { "from": "2a18f49f-67d1-416b-b75d-181b7fbda1d6", "kind": "data_flow", "to": "87a5267d-3694-4e7...
[ { "clean": false, "connect": true, "entity_resolution": true, "leakage_free": true, "model": true, "model_kind": "graph", "serve": true }, { "clean": true, "connect": false, "entity_resolution": true, "leakage_free": true, "model": true, "model_kind": "graph",...
{ "clean": true, "connect": true, "entity_resolution": true, "leakage_free": true, "model": true, "model_kind": "graph", "serve": true }
{ "description": "Score risk across our account/transaction graph using its relationships.", "id": "M-12", "initial_state": {}, "required": { "clean": true, "connect": true, "entity_resolution": true, "leakage_free": true, "model": true, "model_kind": "graph", "serve": true }, "t...
null
{ "execution_graph": { "edges": [ { "from": "a89893d2-c81e-422b-b5d2-134f99f35780", "kind": "data_flow", "to": "6072456e-3352-4b58-8407-b3bdc12186e4" }, { "from": "6072456e-3352-4b58-8407-b3bdc12186e4", "kind": "data_flow", "to": "3adb768a-dffe-46a...
[ { "act": false, "approval_gated": true, "connect": true, "explain": "gradcam", "model": true, "model_kind": "vision", "serve": true }, { "act": true, "approval_gated": false, "connect": true, "explain": "gradcam", "model": true, "model_kind": "vision", "se...
{ "act": true, "approval_gated": true, "connect": true, "explain": "gradcam", "model": true, "model_kind": "vision", "serve": true }
{ "description": "Classify/inspect our images for defects.", "id": "M-13", "initial_state": {}, "required": { "act": true, "approval_gated": true, "connect": true, "explain": "gradcam", "model": true, "model_kind": "vision", "serve": true }, "tool_names": [ "connect", "cl...
null
{ "execution_graph": { "edges": [ { "from": "4c5326d3-65a5-4113-92f2-214d0761252a", "kind": "data_flow", "to": "c1781d97-06ef-4d9d-b975-322fedf4fe62" }, { "from": "c1781d97-06ef-4d9d-b975-322fedf4fe62", "kind": "data_flow", "to": "afb642e9-1dbd-426...
[ { "connect": false, "model": true, "model_kind": "speech", "serve": true }, { "connect": true, "model": false, "model_kind": "speech", "serve": true }, { "connect": true, "model": true, "model_kind": "speech__wrong", "serve": true } ]
{ "connect": true, "model": true, "model_kind": "speech", "serve": true }
{ "description": "Transcribe + analyze our support calls.", "id": "M-14", "initial_state": {}, "required": { "connect": true, "model": true, "model_kind": "speech", "serve": true }, "tool_names": [ "connect", "clean", "model", "index", "serve", "act" ] }
null
{ "execution_graph": { "edges": [ { "from": "5490cd75-1a66-4371-a6e5-e9412993ced4", "kind": "data_flow", "to": "c2c2bbf4-2b63-4a9b-89cb-110dc33850ce" }, { "from": "c2c2bbf4-2b63-4a9b-89cb-110dc33850ce", "kind": "data_flow", "to": "5ca65e14-1ea3-4d3...
[ { "act": false, "approval_gated": true, "connect": true, "exploration": true, "holdout": true, "model": true, "model_kind": "bandit", "serve": true }, { "act": true, "approval_gated": false, "connect": true, "exploration": true, "holdout": true, "model": t...
{ "act": true, "approval_gated": true, "connect": true, "exploration": true, "holdout": true, "model": true, "model_kind": "bandit", "serve": true }
{ "description": "Choose offers online + keep learning which works, safely.", "id": "M-15", "initial_state": {}, "required": { "act": true, "approval_gated": true, "connect": true, "exploration": true, "holdout": true, "model": true, "model_kind": "bandit", "serve": true }, "...
null
{ "execution_graph": { "edges": [ { "from": "c230e242-d9b0-46a0-a692-eb89241bf6c7", "kind": "data_flow", "to": "2a3d963a-1257-4467-9299-f35fd3d7ba4e" }, { "from": "2a3d963a-1257-4467-9299-f35fd3d7ba4e", "kind": "data_flow", "to": "6029f07d-2341-427...
[ { "act": false, "approval_gated": true, "connect": true, "constraints": true, "model": true, "model_kind": "optimization", "serve": true }, { "act": true, "approval_gated": false, "connect": true, "constraints": true, "model": true, "model_kind": "optimization...
{ "act": true, "approval_gated": true, "connect": true, "constraints": true, "model": true, "model_kind": "optimization", "serve": true }
{ "description": "Optimize our plan (routing/allocation) under real constraints.", "id": "M-16", "initial_state": {}, "required": { "act": true, "approval_gated": true, "connect": true, "constraints": true, "model": true, "model_kind": "optimization", "serve": true }, "tool_names...
null
{ "execution_graph": { "edges": [ { "from": "13c6b458-0bed-42b3-bc8e-3381a4fd71d9", "kind": "data_flow", "to": "e26b38ed-0712-4c3f-97e0-bfbaaee1b245" }, { "from": "e26b38ed-0712-4c3f-97e0-bfbaaee1b245", "kind": "data_flow", "to": "238a1921-a210-4f9...
[ { "calibrated": false, "clean": true, "connect": true, "model": true, "model_kind": "bayesian", "serve": true, "uncertainty": true }, { "calibrated": true, "clean": false, "connect": true, "model": true, "model_kind": "bayesian", "serve": true, "uncertaint...
{ "calibrated": true, "clean": true, "connect": true, "model": true, "model_kind": "bayesian", "serve": true, "uncertainty": true }
{ "description": "Give predictions WITH calibrated uncertainty on sparse data.", "id": "M-17", "initial_state": {}, "required": { "calibrated": true, "clean": true, "connect": true, "model": true, "model_kind": "bayesian", "serve": true, "uncertainty": true }, "tool_names": [ ...
null
{ "execution_graph": { "edges": [ { "from": "2767cd8b-e595-4cfd-a7be-07c67cf03ee2", "kind": "data_flow", "to": "30ed92f3-2343-4a21-9e0f-641aec9a5277" }, { "from": "30ed92f3-2343-4a21-9e0f-641aec9a5277", "kind": "data_flow", "to": "ad20b97c-3f8e-47b...
[ { "calibrated": false, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "ensemble", "serve": true, "temporal_validation": true }, { "calibrated": true, "clean": false, "connect": true, "explain": "shap", ...
{ "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "ensemble", "serve": true, "temporal_validation": true }
{ "description": "Squeeze the most accuracy out by combining models, honestly.", "id": "M-18", "initial_state": {}, "required": { "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "ensemble", "serve": true, ...
null
{ "execution_graph": { "edges": [ { "from": "4d170ca8-a59c-4bd3-a3f4-e33231890e1e", "kind": "data_flow", "to": "5096f2af-b3f6-429e-a4d4-83fbb1cf76ec" }, { "from": "5096f2af-b3f6-429e-a4d4-83fbb1cf76ec", "kind": "data_flow", "to": "629801c5-ee8c-42f...
[ { "active_learning": false, "calibrated": true, "clean": true, "connect": true, "label_efficient": true, "model": true, "serve": true }, { "active_learning": true, "calibrated": false, "clean": true, "connect": true, "label_efficient": true, "model": true, ...
{ "active_learning": true, "calibrated": true, "clean": true, "connect": true, "label_efficient": true, "model": true, "serve": true }
{ "description": "Get a good model with very few labels (label-efficient).", "id": "M-19", "initial_state": {}, "required": { "active_learning": true, "calibrated": true, "clean": true, "connect": true, "label_efficient": true, "model": true, "serve": true }, "tool_names": [ ...
null
{ "execution_graph": { "edges": [ { "from": "3a28b633-5a28-4469-a0c8-c39a0d16bd9f", "kind": "data_flow", "to": "4d12d878-9679-4d85-b9de-c609ef08a695" }, { "from": "4d12d878-9679-4d85-b9de-c609ef08a695", "kind": "data_flow", "to": "33570131-8b78-4ce...
[ { "clean": false, "connect": true, "model": true, "privacy_preserving": true, "serve": true, "synthetic_data": true }, { "clean": true, "connect": false, "model": true, "privacy_preserving": true, "serve": true, "synthetic_data": true }, { "clean": true, ...
{ "clean": true, "connect": true, "model": true, "privacy_preserving": true, "serve": true, "synthetic_data": true }
{ "description": "Train on sensitive data without exposing it (privacy-preserving).", "id": "M-20", "initial_state": {}, "required": { "clean": true, "connect": true, "model": true, "privacy_preserving": true, "serve": true, "synthetic_data": true }, "tool_names": [ "connect", ...
null
{ "execution_graph": { "edges": [ { "from": "f089e5be-e188-4e34-9e3f-76b2af0878ff", "kind": "data_flow", "to": "f11d9149-223b-4874-bdc1-782a889c5012" }, { "from": "f11d9149-223b-4874-bdc1-782a889c5012", "kind": "data_flow", "to": "81d02fc5-8d0c-422...
[ { "act": false, "approval_gated": true, "connect": true, "drift_monitoring": true, "model": true, "model_kind": "online", "serve": true, "streaming": true }, { "act": true, "approval_gated": false, "connect": true, "drift_monitoring": true, "model": true, ...
{ "act": true, "approval_gated": true, "connect": true, "drift_monitoring": true, "model": true, "model_kind": "online", "serve": true, "streaming": true }
{ "description": "Keep the model fresh as data streams in, watching for drift.", "id": "M-21", "initial_state": {}, "required": { "act": true, "approval_gated": true, "connect": true, "drift_monitoring": true, "model": true, "model_kind": "online", "serve": true, "streaming": tru...
null
{ "execution_graph": { "edges": [ { "from": "3a110f1e-1fcb-4457-99a8-217d59f1b81b", "kind": "data_flow", "to": "804ecb3f-5a10-468a-a9d4-a24fd29edbb5" }, { "from": "804ecb3f-5a10-468a-a9d4-a24fd29edbb5", "kind": "data_flow", "to": "889aed9d-655f-441...
[ { "baseline_glm": false, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }, { "baseline_glm": true, "calibrated"...
{ "baseline_glm": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Decide credit risk on applicants, defensibly (regulated).", "id": "UC-13", "initial_state": {}, "required": { "baseline_glm": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", ...
null
{ "execution_graph": { "edges": [ { "from": "8f52b9e5-b4b5-42d9-8e23-44cae2335a57", "kind": "data_flow", "to": "b15f9b37-0b17-40dc-8802-3241069ffac5" }, { "from": "b15f9b37-0b17-40dc-8802-3241069ffac5", "kind": "data_flow", "to": "1c1c4451-89d2-439...
[ { "act": false, "approval_gated": true, "connect": true, "constraints": true, "model": true, "model_kind": "optimization", "serve": true }, { "act": true, "approval_gated": false, "connect": true, "constraints": true, "model": true, "model_kind": "optimization...
{ "act": true, "approval_gated": true, "connect": true, "constraints": true, "model": true, "model_kind": "optimization", "serve": true }
{ "description": "Optimize our prices/discounts within guardrails.", "id": "UC-14", "initial_state": {}, "required": { "act": true, "approval_gated": true, "connect": true, "constraints": true, "model": true, "model_kind": "optimization", "serve": true }, "tool_names": [ "con...
null
{ "execution_graph": { "edges": [ { "from": "0121ac1c-200c-426f-a07c-09ea483daff8", "kind": "data_flow", "to": "5833962e-9b1d-40ad-9d9b-7809e9fb334f" }, { "from": "5833962e-9b1d-40ad-9d9b-7809e9fb334f", "kind": "data_flow", "to": "9d686f90-d2fb-4d8...
[ { "calibrated": false, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }, { "calibrated": true, "clean": false, "connect": true, ...
{ "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Predict which employees are at risk of leaving.", "id": "UC-15", "initial_state": {}, "required": { "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serv...
null
{ "execution_graph": { "edges": [ { "from": "3e204333-29b0-4232-9713-ef5967cb8562", "kind": "data_flow", "to": "7db9d64d-1815-4c9f-8fed-bab9fd24592e" }, { "from": "7db9d64d-1815-4c9f-8fed-bab9fd24592e", "kind": "data_flow", "to": "4b41ce8b-3b03-4fb...
[ { "act": false, "approval_gated": true, "clean": true, "connect": true, "grounded_in_data": true, "model": true, "model_kind": "classifier", "serve": true }, { "act": true, "approval_gated": false, "clean": true, "connect": true, "grounded_in_data": true, ...
{ "act": true, "approval_gated": true, "clean": true, "connect": true, "grounded_in_data": true, "model": true, "model_kind": "classifier", "serve": true }
{ "description": "Triage + route incoming support tickets automatically.", "id": "UC-16", "initial_state": {}, "required": { "act": true, "approval_gated": true, "clean": true, "connect": true, "grounded_in_data": true, "model": true, "model_kind": "classifier", "serve": true }...
null
{ "execution_graph": { "edges": [ { "from": "966fcf6c-1dd1-4143-88e8-ae4ec6816571", "kind": "data_flow", "to": "6362a584-5718-4acb-b10d-b420f828fc68" }, { "from": "6362a584-5718-4acb-b10d-b420f828fc68", "kind": "data_flow", "to": "8fd22084-f3e9-432...
[ { "act": false, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "monitoring": true, "point_in_time": true, "serve": true, "temporal_validation": true }, {...
{ "act": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "monitoring": true, "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Monitor our suppliers for emerging risk.", "id": "UC-18", "initial_state": {}, "required": { "act": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt"...
null
{ "execution_graph": { "edges": [ { "from": "fdc842d4-3515-4597-a1d7-919bc011be9e", "kind": "data_flow", "to": "b71c7691-985d-42f0-bedb-ee44069d22a3" }, { "from": "b71c7691-985d-42f0-bedb-ee44069d22a3", "kind": "data_flow", "to": "197ef6ff-e6c8-458...
[ { "act": false, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }, { "act": true, "a...
{ "act": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "point_in_time": true, "serve": true, "temporal_validation": true }
{ "description": "Predict claim denials + patient no-shows.", "id": "UC-19", "initial_state": {}, "required": { "act": true, "approval_gated": true, "calibrated": true, "clean": true, "connect": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt...
null
{ "execution_graph": { "edges": [ { "from": "81e98908-f9db-4450-972d-c673f1c51d11", "kind": "data_flow", "to": "c909244d-c876-4c52-a0bb-09941faa7d82" }, { "from": "c909244d-c876-4c52-a0bb-09941faa7d82", "kind": "data_flow", "to": "beb66c58-1f5a-4b7...
[ { "backtest": false, "connect": true, "model": true, "model_kind": "forecast", "serve": true, "temporal_validation": true }, { "backtest": true, "connect": false, "model": true, "model_kind": "forecast", "serve": true, "temporal_validation": true }, { "bac...
{ "backtest": true, "connect": true, "model": true, "model_kind": "forecast", "serve": true, "temporal_validation": true }
{ "description": "Forecast cash flow + working capital.", "id": "UC-20", "initial_state": {}, "required": { "backtest": true, "connect": true, "model": true, "model_kind": "forecast", "serve": true, "temporal_validation": true }, "tool_names": [ "connect", "clean", "model...
null
{ "execution_graph": { "edges": [ { "from": "78a0943e-fd03-44e0-ada6-2ad0b91eece0", "kind": "data_flow", "to": "bb85db2d-7f09-4499-9302-61c471b6dd32" }, { "from": "bb85db2d-7f09-4499-9302-61c471b6dd32", "kind": "data_flow", "to": "2a2aeab9-715e-404...
[ { "act": false, "connect": true, "drift_monitoring": true, "drivers": true, "model": true, "model_kind": "anomaly", "serve": true, "streaming": true }, { "act": true, "connect": false, "drift_monitoring": true, "drivers": true, "model": true, "model_kind":...
{ "act": true, "connect": true, "drift_monitoring": true, "drivers": true, "model": true, "model_kind": "anomaly", "serve": true, "streaming": true }
{ "description": "Detect + explain incidents across our systems (AIOps).", "id": "UC-21", "initial_state": {}, "required": { "act": true, "connect": true, "drift_monitoring": true, "drivers": true, "model": true, "model_kind": "anomaly", "serve": true, "streaming": true }, "t...
null
{ "execution_graph": { "edges": [ { "from": "d5db4312-3a50-4277-975f-20356385cdfe", "kind": "data_flow", "to": "5d98e89d-8e4d-4229-a89f-36951124735b" }, { "from": "5d98e89d-8e4d-4229-a89f-36951124735b", "kind": "data_flow", "to": "44dd5665-f3ab-478...
[ { "clean": false, "connect": true, "drivers": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "serve": true }, { "clean": true, "connect": false, "drivers": true, "explain": "shap", "leakage_free": true, "model": true, ...
{ "clean": true, "connect": true, "drivers": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "serve": true }
{ "description": "Find what drives scrap/quality/yield in our process.", "id": "UC-22", "initial_state": {}, "required": { "clean": true, "connect": true, "drivers": true, "explain": "shap", "leakage_free": true, "model": true, "model_kind": "gbdt", "serve": true }, "tool_nam...
null
{ "execution_graph": { "edges": [ { "from": "cba98d71-1c44-489f-b722-e24b03429a7f", "kind": "data_flow", "to": "31f47263-97d2-4c18-bf63-00d6ce62833c" }, { "from": "31f47263-97d2-4c18-bf63-00d6ce62833c", "kind": "data_flow", "to": "b65d2c19-c8cd-47e...
[ { "act": false, "approval_gated": true, "clean": true, "connect": true, "holdout": true, "incrementality": true, "leakage_free": true, "model": true, "model_kind": "uplift", "serve": true }, { "act": true, "approval_gated": false, "clean": true, "connect":...
{ "act": true, "approval_gated": true, "clean": true, "connect": true, "holdout": true, "incrementality": true, "leakage_free": true, "model": true, "model_kind": "uplift", "serve": true }
{ "description": "Target our campaigns at who'll actually respond because of them.", "id": "UC-23", "initial_state": {}, "required": { "act": true, "approval_gated": true, "clean": true, "connect": true, "holdout": true, "incrementality": true, "leakage_free": true, "model": true...
null
{ "execution_graph": { "edges": [ { "from": "c365ee19-1ed7-43b1-94a6-f3ce89b7e1e7", "kind": "data_flow", "to": "60f97b33-2bbe-40f1-a747-c6b840fa502c" }, { "from": "60f97b33-2bbe-40f1-a747-c6b840fa502c", "kind": "data_flow", "to": "4492cb65-3167-4f5...
[ { "act": false, "approval_gated": true, "clean": true, "connect": true, "extraction": true, "grounded_in_data": true, "index": true, "index_kind": "semantic", "model": true, "model_kind": "document_intelligence", "serve": true }, { "act": true, "approval_gated...
{ "act": true, "approval_gated": true, "clean": true, "connect": true, "extraction": true, "grounded_in_data": true, "index": true, "index_kind": "semantic", "model": true, "model_kind": "document_intelligence", "serve": true }
{ "description": "Read our contracts + flag obligation/renewal risk.", "id": "UC-24", "initial_state": {}, "required": { "act": true, "approval_gated": true, "clean": true, "connect": true, "extraction": true, "grounded_in_data": true, "index": true, "index_kind": "semantic", ...
null
{ "execution_graph": { "edges": [ { "from": "67c7b337-89be-489d-bff3-c4dded6bef4d", "kind": "data_flow", "to": "46d005d1-aabb-46ac-99ab-00b58f140c5f" }, { "from": "46d005d1-aabb-46ac-99ab-00b58f140c5f", "kind": "data_flow", "to": "f1de50e4-e3a5-48e...
[ { "clean": false, "connect": true, "entity_resolution": true, "holdout": true, "labels": true, "leakage_free": true, "point_in_time": true }, { "clean": true, "connect": false, "entity_resolution": true, "holdout": true, "labels": true, "leakage_free": true, ...
{ "clean": true, "connect": true, "entity_resolution": true, "holdout": true, "labels": true, "leakage_free": true, "point_in_time": true }
{ "description": "Connect our systems, resolve entities, and build a leakage-safe history to learn from — with holdouts kept sacred.", "id": "DSC", "initial_state": {}, "required": { "clean": true, "connect": true, "entity_resolution": true, "holdout": true, "labels": true, "leakage_free...
null
{ "execution_graph": { "edges": [ { "from": "47866639-5e04-4ff3-8224-a749e6c041d7", "kind": "data_flow", "to": "abe79332-355c-47d8-8ed2-7531f3409afe" }, { "from": "abe79332-355c-47d8-8ed2-7531f3409afe", "kind": "data_flow", "to": "74883853-ffa9-499...
[ { "act": false, "approval_gated": true, "connect": true, "model": true, "serve": true }, { "act": true, "approval_gated": false, "connect": true, "model": true, "serve": true }, { "act": true, "approval_gated": true, "connect": false, "model": true, ...
{ "act": true, "approval_gated": true, "connect": true, "model": true, "serve": true }
{ "description": "Let it observe + recommend, but require approval before it acts on the world.", "id": "AC", "initial_state": {}, "required": { "act": true, "approval_gated": true, "connect": true, "model": true, "serve": true }, "tool_names": [ "connect", "clean", "model", ...
null
{ "execution_graph": { "edges": [ { "from": "40583423-b7fe-44be-bc22-2bfc9a0cad5b", "kind": "data_flow", "to": "283bfd16-c3fd-4abc-ad8d-cbc732f04e62" } ], "nodes": [ { "id": "40583423-b7fe-44be-bc22-2bfc9a0cad5b", "kind": "connect", "name":...
[ { "abstain": false, "in_scope": false }, { "abstain": true, "in_scope": true } ]
{ "abstain": true, "in_scope": false }
{ "description": "Write me a poem about our quarterly numbers.", "id": "ABSTAIN", "initial_state": {}, "required": { "abstain": true, "in_scope": false }, "tool_names": [ "connect", "clean", "model", "index", "serve", "act" ] }
null
{ "execution_graph": { "edges": [ { "from": "068e8543-05e3-431f-a45b-aaf35a3e1879", "kind": "data_flow", "to": "5270d8f3-139f-44ec-9b54-7dd564492293" } ], "nodes": [ { "id": "068e8543-05e3-431f-a45b-aaf35a3e1879", "kind": "connect", "name":...
[ { "abstain": false, "in_scope": false }, { "abstain": true, "in_scope": true } ]
{ "abstain": true, "in_scope": false }
{ "description": "What's the capital of France?", "id": "ABSTAIN", "initial_state": {}, "required": { "abstain": true, "in_scope": false }, "tool_names": [ "connect", "clean", "model", "index", "serve", "act" ] }
null
{ "execution_graph": { "edges": [ { "from": "7adf3a52-ac51-4cf9-93f1-cba7c79e73cc", "kind": "data_flow", "to": "4c98c149-ddd9-4d13-abc0-f50902640ead" } ], "nodes": [ { "id": "7adf3a52-ac51-4cf9-93f1-cba7c79e73cc", "kind": "connect", "name":...
[ { "abstain": false, "in_scope": false }, { "abstain": true, "in_scope": true } ]
{ "abstain": true, "in_scope": false }
{ "description": "Book me a flight to Tokyo next week.", "id": "ABSTAIN", "initial_state": {}, "required": { "abstain": true, "in_scope": false }, "tool_names": [ "connect", "clean", "model", "index", "serve", "act" ] }
null
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **failed_payments** at 4800.0% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer", "numbers_fro...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "What's driving churn in our accounts?", "id": "qa_results/0/0/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "index", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **failed_payments** at 4800.0% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer", "numbers_fro...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Which factors matter most here?", "id": "qa_results/0/0/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "index", "ser...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **failed_payments** at 4800.0% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer", "numbers_fro...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "What are the top drivers of this model?", "id": "qa_results/0/0/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "index", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **failed_payments** at 4800.0% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer", "numbers_fro...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Why is the model flagging accounts?", "id": "qa_results/0/0/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "index", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **ACME-1000** at risk 0.940, driven mainly by failed_payments.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, ...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }
{ "description": "Which accounts are most at risk?", "id": "qa_results/0/1/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **ACME-1000** at risk 0.940, driven mainly by failed_payments.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, ...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }
{ "description": "Who's most likely to churn?", "id": "qa_results/0/1/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "m...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **ACME-1000** at risk 0.940, driven mainly by failed_payments.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, ...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }
{ "description": "Show me the riskiest accounts.", "id": "qa_results/0/1/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **ACME-1000** at risk 0.940, driven mainly by failed_payments.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, ...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }
{ "description": "Which customers should we worry about?", "id": "qa_results/0/1/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, "tool_names": [ "connect", "cle...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "How good is this model?", "id": "qa_results/0/2/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Is this model any good?", "id": "qa_results/0/2/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "What's the model's accuracy?", "id": "qa_results/0/2/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model",...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "How well does this perform?", "id": "qa_results/0/2/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "What does the AUC mean here?", "id": "qa_results/0/3/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model",...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Explain the measure to me.", "id": "qa_results/0/3/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Is that AUC good or bad?", "id": "qa_results/0/3/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "How should I read this score?", "id": "qa_results/0/3/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model"...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Is this model overfit?", "id": "qa_results/0/4/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Should I trust these numbers?", "id": "qa_results/0/4/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model"...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Does this look too good to be true?", "id": "qa_results/0/4/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Any sign of leakage?", "id": "qa_results/0/4/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "in...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **failed_payments** at 4800.0% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer", "numbers_fro...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "What's the single biggest driver?", "id": "qa_results/0/5/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "index", "s...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **failed_payments** at 4800.0% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer", "numbers_fro...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Which feature dominates?", "id": "qa_results/0/5/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "index", "serve", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **failed_payments** at 4800.0% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer", "numbers_fro...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "What's the number one factor?", "id": "qa_results/0/5/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "index", "serve...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **failed_payments** at 4800.0% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer", "numbers_fro...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Which driver carries the most weight?", "id": "qa_results/0/5/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", "index", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Is the model calibrated?", "id": "qa_results/0/6/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Are these probabilities real probabilities?", "id": "qa_results/0/6/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "Can I treat the risk scores as likelihoods?", "id": "qa_results/0/6/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.720** (accuracy 0.660), with failed_payments the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, { "groun...
{ "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }
{ "description": "How are the scores calibrated?", "id": "qa_results/0/6/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.72, "numbers_from_context": true, "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", "model...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **ACME-1000** at risk 0.940, driven mainly by failed_payments.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, ...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }
{ "description": "Give me the top account and why.", "id": "qa_results/0/7/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, "tool_names": [ "connect", "clean", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **ACME-1000** at risk 0.940, driven mainly by failed_payments.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, ...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }
{ "description": "Who is the highest risk and what's driving it?", "id": "qa_results/0/7/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, "tool_names": [ "connect", ...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **ACME-1000** at risk 0.940, driven mainly by failed_payments.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, ...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }
{ "description": "Name the riskiest account and the reason.", "id": "qa_results/0/7/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, "tool_names": [ "connect", "...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **ACME-1000** at risk 0.940, driven mainly by failed_payments.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, ...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }
{ "description": "Which account tops the list, and why is it there?", "id": "qa_results/0/7/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "ACME-1000", "top_driver": "failed_payments" }, "tool_names": [ "connect...
Project: Lead scoring - inbound #1 Model: gbdt, trained on leads. Measure: AUC 0.720, accuracy 0.660. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): failed_payments: 4800.0% tenure_days: 2400.0% last_login_days: 1600.0% contract_mo...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **support_severity** at 3952.1% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer", "numbers_f...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "What's driving churn in our accounts?", "id": "qa_results/1/0/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", "index", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **support_severity** at 3952.1% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer", "numbers_f...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Which factors matter most here?", "id": "qa_results/1/0/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", "index", "se...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **support_severity** at 3952.1% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer", "numbers_f...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "What are the top drivers of this model?", "id": "qa_results/1/0/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", "index",...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **support_severity** at 3952.1% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer", "numbers_f...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Why is the model flagging accounts?", "id": "qa_results/1/0/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", "index", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **NORTHWIND-1013** at risk 0.929, driven mainly by support_severity.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_sev...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }
{ "description": "Which accounts are most at risk?", "id": "qa_results/1/1/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }, "tool_names": [ "connect", "cle...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **NORTHWIND-1013** at risk 0.929, driven mainly by support_severity.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_sev...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }
{ "description": "Who's most likely to churn?", "id": "qa_results/1/1/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **NORTHWIND-1013** at risk 0.929, driven mainly by support_severity.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_sev...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }
{ "description": "Show me the riskiest accounts.", "id": "qa_results/1/1/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }, "tool_names": [ "connect", "clean...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The highest-risk account is **NORTHWIND-1013** at risk 0.929, driven mainly by support_severity.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_sev...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }
{ "description": "Which customers should we worry about?", "id": "qa_results/1/1/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_account": "NORTHWIND-1013", "top_driver": "support_severity" }, "tool_names": [ "connect", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "How good is this model?", "id": "qa_results/1/2/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Is this model any good?", "id": "qa_results/1/2/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "What's the model's accuracy?", "id": "qa_results/1/2/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model"...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "How well does this perform?", "id": "qa_results/1/2/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model",...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "What does the AUC mean here?", "id": "qa_results/1/3/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model"...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Explain the measure to me.", "id": "qa_results/1/3/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Is that AUC good or bad?", "id": "qa_results/1/3/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "How should I read this score?", "id": "qa_results/1/3/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Is this model overfit?", "id": "qa_results/1/4/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Should I trust these numbers?", "id": "qa_results/1/4/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Does this look too good to be true?", "id": "qa_results/1/4/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Any sign of leakage?", "id": "qa_results/1/4/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", "i...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **support_severity** at 3952.1% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer", "numbers_f...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "What's the single biggest driver?", "id": "qa_results/1/5/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", "index", "...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **support_severity** at 3952.1% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer", "numbers_f...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Which feature dominates?", "id": "qa_results/1/5/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", "index", "serve", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **support_severity** at 3952.1% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer", "numbers_f...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "What's the number one factor?", "id": "qa_results/1/5/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", "index", "serv...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "The strongest driver is **support_severity** at 3952.1% of total feature importance.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer", "numbers_f...
{ "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Which driver carries the most weight?", "id": "qa_results/1/5/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", "index", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Is the model calibrated?", "id": "qa_results/1/6/0", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "model", ...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Are these probabilities real probabilities?", "id": "qa_results/1/6/1", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clea...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "Can I treat the risk scores as likelihoods?", "id": "qa_results/1/6/2", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clea...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
{ "execution_graph": { "edges": [], "nodes": [] }, "kind": "answer", "message": "This model scores **AUC 0.790** (accuracy 0.710), with support_severity the strongest driver.", "tool_calls": [] }
[ { "grounded": false, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "grounded": true, "kind": "answer__wrong", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, { "gro...
{ "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }
{ "description": "How are the scores calibrated?", "id": "qa_results/1/6/3", "initial_state": {}, "required": { "grounded": true, "kind": "answer", "measure_auc": 0.79, "numbers_from_context": true, "top_driver": "support_severity" }, "tool_names": [ "connect", "clean", "mode...
Project: Lifetime value forecast #2 Model: regression, trained on customers. Measure: AUC 0.790, accuracy 0.710. Calibrated: yes. Drivers, strongest first - each shown as its share of the model's total feature importance (they sum to 100%): support_severity: 3952.1% total_events: 2155.7% ticket_count: 1556.9% u...
End of preview.

base-agent-interaction-corpus — v16

Training corpus for the Orama Studio base agent. Private. Internal-confidential.

Training cases 16,734
Held-out canary 1,794
Cases carrying conversation history 855 (5.1%)
Generator zetta::dataset::BaseAgentDatasetGenerator, commit c13070a
Admission-gated yes — every case passes admit_corpus before it is written

What it is

Deterministically synthesized from the solutions catalogue: 18 interaction families crossed with synthetic Project contexts and hand-written paraphrases. Each case carries the ask, the retrieved grounding, the reference response, the facts a correct response must satisfy, and negative controls that must fail — discrimination is the product, not the pass rate.

Limitations — read these

This corpus has not produced a good agent, and four checkpoints say so.

  • Multi-turn density is 5.1%. At the previous 2.2%, three checkpoints (v13/v14/v15) all answered two different asks with a byte-identical reply. Raising it to 5.1% fixed that symptom in v16 — and v16 is still not good: it answers a project request by describing what Studio can do, follows "yes" with an unprompted non sequitur, and fabricates claims.
  • A family's register is load-bearing, and nothing checks it. In v14 the redirect reply opened with a phrase appearing nowhere else in 16,239 cases; the model applied it to substantive requests it was never meant for, regressing behaviour v13 had right. 180 cases (1.1%) caused it. Fixed by rephrasing, but no invariant catches the class.
  • Hand-authored coverage has a ceiling. Four rounds produced four narrow fixes and no broad gain. A hand-written corpus can only encode failures someone already observed. The intended successor is the RLVR loop (zetta#859), which generates observation-conditioned experience rather than guessing at it.
  • Product claims are checked but the registry is small. ProductClaimsAreTrue caught four cases asserting "You can run Python in a Pipeline step" when the engine has only simulated Python. The registry covers a handful of claims, not all of them.

Canary

canary.jsonl holds whole Projects the training split never sees, so grading it is a real memorization check rather than a resample.

Provenance

Reports: docs/reports/base-agent-validation-2026-08-06.md in oramasearch/zetta. Checkpoints trained on this lineage: OramaSearch/base-agent-0.6b.

Downloads last month
6