Datasets:

getomnico
/

omni-enterprise-rag-bench

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
selection_rule: string
count: int64
question_ids: list<item: string>
  child 0, item: string
by_type: struct<basic: int64, semantic: int64, intra_document_reasoning: int64, project_related: int64, const (... 35 chars omitted)
  child 0, basic: int64
  child 1, semantic: int64
  child 2, intra_document_reasoning: int64
  child 3, project_related: int64
  child 4, constrained: int64
  child 5, completeness: int64
eval_summary: struct<completed_utc: timestamp[s], patch3_result: string, gpt54_accepted_ids: list<item: string>, p (... 783 chars omitted)
  child 0, completed_utc: timestamp[s]
  child 1, patch3_result: string
  child 2, gpt54_accepted_ids: list<item: string>
      child 0, item: string
  child 3, p3_eval_stats: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
      child 0, n: int64
      child 1, correctness: double
      child 2, completeness: double
      child 3, overall: double
      child 4, recall: double
      child 5, invalid: double
  child 4, base_before: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
      child 0, n: int64
      child 1, correctness: double
      child 2, completeness: double
      child 3, overall: double
      child 4, recall: double
      child 5, invalid: double
  child 5, base_plus_patch3_selective: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
      child 0, n: int64
      child 1, correctness: double
      child 2, completeness: double
      child 3, overall: double
      child 4, recall: double
      child 5, invalid: double
  child 6, patch3_ids_selected_on_base: list<item: string>
      child 0, item: string
  child 7, base_plus_patch12_selective: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
      child 0, n: int64
      child 1, correctness: double
      child 2, completeness: double
      child 3, overall: double
      child 4, recall: double
      child 5, invalid: double
  child 8, base_plus_patch12_plus_patch3_selective: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
      child 0, n: int64
      child 1, correctness: double
      child 2, completeness: double
      child 3, overall: double
      child 4, recall: double
      child 5, invalid: double
  child 9, patch3_ids_selected_after_patch12: list<item: string>
      child 0, item: string
question_id: string
document_ids: list<item: string>
  child 0, item: string
answer: string
to
{'question_id': Value('string'), 'answer': Value('string'), 'document_ids': List(Value('string'))}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 99, in get_rows_or_raise
                  return get_rows(
                         ^^^^^^^^^
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 77, in get_rows
                  rows_plus_one = list(itertools.islice(ds, rows_max_number + 1))
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2690, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2227, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2251, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 494, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 384, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 299, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 128, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2321, in table_cast
                  return cast_table_to_schema(table, schema)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2249, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              selection_rule: string
              count: int64
              question_ids: list<item: string>
                child 0, item: string
              by_type: struct<basic: int64, semantic: int64, intra_document_reasoning: int64, project_related: int64, const (... 35 chars omitted)
                child 0, basic: int64
                child 1, semantic: int64
                child 2, intra_document_reasoning: int64
                child 3, project_related: int64
                child 4, constrained: int64
                child 5, completeness: int64
              eval_summary: struct<completed_utc: timestamp[s], patch3_result: string, gpt54_accepted_ids: list<item: string>, p (... 783 chars omitted)
                child 0, completed_utc: timestamp[s]
                child 1, patch3_result: string
                child 2, gpt54_accepted_ids: list<item: string>
                    child 0, item: string
                child 3, p3_eval_stats: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
                    child 0, n: int64
                    child 1, correctness: double
                    child 2, completeness: double
                    child 3, overall: double
                    child 4, recall: double
                    child 5, invalid: double
                child 4, base_before: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
                    child 0, n: int64
                    child 1, correctness: double
                    child 2, completeness: double
                    child 3, overall: double
                    child 4, recall: double
                    child 5, invalid: double
                child 5, base_plus_patch3_selective: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
                    child 0, n: int64
                    child 1, correctness: double
                    child 2, completeness: double
                    child 3, overall: double
                    child 4, recall: double
                    child 5, invalid: double
                child 6, patch3_ids_selected_on_base: list<item: string>
                    child 0, item: string
                child 7, base_plus_patch12_selective: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
                    child 0, n: int64
                    child 1, correctness: double
                    child 2, completeness: double
                    child 3, overall: double
                    child 4, recall: double
                    child 5, invalid: double
                child 8, base_plus_patch12_plus_patch3_selective: struct<n: int64, correctness: double, completeness: double, overall: double, recall: double, invalid (... 9 chars omitted)
                    child 0, n: int64
                    child 1, correctness: double
                    child 2, completeness: double
                    child 3, overall: double
                    child 4, recall: double
                    child 5, invalid: double
                child 9, patch3_ids_selected_after_patch12: list<item: string>
                    child 0, item: string
              question_id: string
              document_ids: list<item: string>
                child 0, item: string
              answer: string
              to
              {'question_id': Value('string'), 'answer': Value('string'), 'document_ids': List(Value('string'))}
              because column names don't match

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.

Omni EnterpriseRAG-Bench Results

This dataset contains the public verification artifacts for Omni's EnterpriseRAG-Bench full-500 agentic benchmark run.

The benchmark itself is EnterpriseRAG-Bench. This repository does not re-host the full benchmark corpus. It publishes Omni's answer files, evaluation result files, patch manifests, and the merge script needed to verify the reported metrics against the benchmark questions and gold metadata.

base/: the clean 500-question DeepSeek V4 Pro agentic run.
final/: the final merged result after the documented GPT-5.4 adjustment pass.
patches/: the small GPT-5.4 remediation and re-judgement artifacts used by final/compute_merged_gpt54_adjusted.py.
metadata/: run metadata and dataset/code provenance.
SHA256SUMS: checksums for all published files.

Reported Final Metrics

From final/results_omni_agentic_deepseek_v4_pro_full500_merged_gpt54_adjusted.json:

Metric	Value
Questions	500
Overall score	68.07302
Correctness	73.4%
Completeness	72.27232%
Document recall	69.43328%
Invalid extra docs	1.27234

Base Run Metrics

From base/results_omni_agentic_deepseek_v4_pro_full500.json:

Metric	Value
Questions	500
Overall score	63.95508
Correctness	69.0%
Completeness	70.47468%
Document recall	69.43328%
Invalid extra docs	1.27234

Adjustment Methodology

The base run used DeepSeek V4 Pro as the agent and judge for all 500 questions. We then applied a small GPT-5.4 high-reasoning adjustment pass to reduce answer-generation and judge-model artifacts while keeping the retrieval evidence unchanged.

The final merge changes 23 of 500 rows:

21 answer rows selected from GPT-5.4 remediation patches where answer contribution improved.
2 judgement-only rows selected from GPT-5.4 re-judgement of suspicious high-completeness failures.
0 submitted document ID rows changed. Retrieval recall is therefore unchanged from the base DeepSeek run.

See final/merge_manifest_gpt54_adjusted.json and final/compute_merged_gpt54_adjusted.py for exact row-level provenance and merge logic.

Recomputing Metrics

The final merge can be recomputed without any LLM calls from the artifact root:

python3 final/compute_merged_gpt54_adjusted.py
cat final/results_omni_agentic_deepseek_v4_pro_full500_merged_gpt54_adjusted.json | jq .aggregate_stats
cat base/results_omni_agentic_deepseek_v4_pro_full500.json | jq .aggregate_stats

The merge script reads base/ and patches/, writes final/, and preserves the base DeepSeek document IDs and retrieval metrics. The metadata/dataset_reference.json file records the EnterpriseRAG-Bench commit and local questions.jsonl checksum used for this run.

Exclusions

This artifact intentionally excludes raw chat traces, model usage records, database dumps, API keys, the full corpus, and embedding vectors. Those are not needed to verify the published answer/evaluation metrics.

Downloads last month: 16

Total file size:

2.57 MB