Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
schema: string
publication_content: string
author_identity: string
projection_id: string
publication_title: string
primary_file: string
coverage_cutoff: timestamp[s]
claim_refs: list<item: null>
  child 0, item: null
claim_states: list<item: null>
  child 0, item: null
repository_visibility_managed_externally: bool
capabilities_removed: list<item: null>
  child 0, item: null
publication_state_sha256: string
operator_role_statement: string
status: string
paper_title: string
authorship_signature_sha256: string
human_authors: list<item: null>
  child 0, item: null
production_mode: string
signature_scheme: string
authorship_statement: string
author: string
paper_sha256: string
production_disclosure: string
authorship_kind: string
paper_file: string
human_operator: struct<basic_logical_semantic_proofreading: bool, domain_knowledge_contributed: bool, external_publi (... 97 chars omitted)
  child 0, basic_logical_semantic_proofreading: bool
  child 1, domain_knowledge_contributed: bool
  child 2, external_public_action_authority: bool
  child 3, initial_high_level_goal: bool
  child 4, substantive_research_contribution: bool
title_filename_required: bool
coding_agent: string
signed_by: string
to
{'author': Value('string'), 'authorship_kind': Value('string'), 'authorship_signature_sha256': Value('string'), 'authorship_statement': Value('string'), 'coding_agent': Value('string'), 'human_authors': List(Value('null')), 'human_operator': {'basic_logical_semantic_proofreading': Value('bool'), 'domain_knowledge_contributed': Value('bool'), 'external_public_action_authority': Value('bool'), 'initial_high_level_goal': Value('bool'), 'substantive_research_contribution': Value('bool')}, 'operator_role_statement': Value('string'), 'paper_file': Value('string'), 'paper_sha256': Value('string'), 'paper_title': Value('string'), 'production_disclosure': Value('string'), 'production_mode': Value('string'), 'publication_content': Value('string'), 'schema': Value('string'), 'signature_scheme': Value('string'), 'signed_by': Value('string'), 'status': Value('string'), 'title_filename_required': Value('bool')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                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 129, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                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
              schema: string
              publication_content: string
              author_identity: string
              projection_id: string
              publication_title: string
              primary_file: string
              coverage_cutoff: timestamp[s]
              claim_refs: list<item: null>
                child 0, item: null
              claim_states: list<item: null>
                child 0, item: null
              repository_visibility_managed_externally: bool
              capabilities_removed: list<item: null>
                child 0, item: null
              publication_state_sha256: string
              operator_role_statement: string
              status: string
              paper_title: string
              authorship_signature_sha256: string
              human_authors: list<item: null>
                child 0, item: null
              production_mode: string
              signature_scheme: string
              authorship_statement: string
              author: string
              paper_sha256: string
              production_disclosure: string
              authorship_kind: string
              paper_file: string
              human_operator: struct<basic_logical_semantic_proofreading: bool, domain_knowledge_contributed: bool, external_publi (... 97 chars omitted)
                child 0, basic_logical_semantic_proofreading: bool
                child 1, domain_knowledge_contributed: bool
                child 2, external_public_action_authority: bool
                child 3, initial_high_level_goal: bool
                child 4, substantive_research_contribution: bool
              title_filename_required: bool
              coding_agent: string
              signed_by: string
              to
              {'author': Value('string'), 'authorship_kind': Value('string'), 'authorship_signature_sha256': Value('string'), 'authorship_statement': Value('string'), 'coding_agent': Value('string'), 'human_authors': List(Value('null')), 'human_operator': {'basic_logical_semantic_proofreading': Value('bool'), 'domain_knowledge_contributed': Value('bool'), 'external_public_action_authority': Value('bool'), 'initial_high_level_goal': Value('bool'), 'substantive_research_contribution': Value('bool')}, 'operator_role_statement': Value('string'), 'paper_file': Value('string'), 'paper_sha256': Value('string'), 'paper_title': Value('string'), 'production_disclosure': Value('string'), 'production_mode': Value('string'), 'publication_content': Value('string'), 'schema': Value('string'), 'signature_scheme': Value('string'), 'signed_by': Value('string'), 'status': Value('string'), 'title_filename_required': Value('bool')}
              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.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

The Iran War Receipt

This directory is one dataset with stable identity the-iran-war-receipt. It is not a series of replacement datasets. dataset.json is the current materialized view; changes.jsonl is the append-only, hash-chained change log; and snapshots.jsonl preserves historical aggregate views. Current dataset revision: 31 Government, military, corporate, NGO, academic, journalistic, and social-media artifacts are stored as sourced claims. No source certifies its own claim. Requested, authorized, obligated, outlayed, consumed, and replacement-liability amounts are not interchangeable. Null amounts are intentionally unknown.

Current CBO military-cost view

  • A recovered, SHA-256-pinned archived copy of CBO report 62756 supports a traceable source-claim view of $38,100,000,000.00 through August 1, 2026: $21.7B munitions, $10.4B flying hours, $2.7B fuel, $1.5B other operations, and a preferred $1,882,000,000.00 equipment-loss estimate. The equipment replace-all counterfactual is $3,312,000,000.00.
  • CBO's equipment table identifies 42 aircraft or uncrewed-aircraft battle losses plus one THAAD radar. Counts, replacement assumptions, and later disposition reporting remain independently contestable; the archived PDF proves what CBO published, not that every modeled input is true.
  • CBO classified $42.3B of a $67.1B Administration DoD request as apparently war-related. Requested budget authority is not incurred expenditure and is excluded from the minimum directly evidenced fiscal-cost aggregate.
  • CBO could not estimate base repair or rebuilding. That liability remains unknown rather than zero.
  • A 12-site USAspending search found a $1,200,000.00 obligation for rapid damage assessment and a restoration/recovery plan at Naval Support Activity Bahrain. The firm-fixed-price NAVFAC engineering award was signed April 10, 2026; USAspending reports no outlay and no national-interest-action tag. The description does not name Iran or the operation, so this is published as a strongly attributable candidate—not silently added to the strict directly evidenced fiscal floor. The other 53 returned award rows are retained as screened non-confirmations unless their scopes can be tied to conflict damage.

Aircraft and UAS loss ledger

  • The event ledger records 63 loss-type aircraft/UAS, 17 damage-only events, and one unresolved KC-135 disposition.
  • Every one of those 17 event groups has exactly one asset_valuations row. Nine loss groups and four damage groups have reproducible conditional proxy amounts; MQ-1, A-10 damage, E-3 damage, and the unresolved KC-135 remain explicitly unvalued. Event rows expose the formula, valuation input IDs, assumptions, uncertainty propagation, containing cost record, and unresolved evidence rather than hiding the arithmetic in an aggregate.
  • Replacement proxies now cover 60 of the 63 loss-type units at $2,800,566,000.00 to $4,081,800,000.00. The remaining three are unresolved MQ-1 variants and stay unvalued.
  • A separate 15-aircraft partial repair proxy is $534,738,000.00 to $1,838,700,000.00; the damaged A-10 and E-3 remain unvalued.
  • These are mixed-concept replacement/repair proxies, not expenditure. CBO and Pentagon equipment totals may already contain many of them, and the overlap graph prevents additive use. CBO labels four losses MH-6 while CRS labels four AH-6; the variant disagreement remains visible.

Modeled base-site reconstruction ledger

These 12 rows reproduce the captured CSIS site table. They are one modeled lineage, not 12 independent confirmations or observed obligations. Each row is structurally contained in the assessed-set total; summing the rows is a reproduction check, not an additional cost.

Site Low Central High
Al Dhafra Air Base $143.4M $360.1M $576.9M
Al Sader Air Base $11.2M $32.9M $54.5M
Al Udeid Air Base $171.9M $318.9M $465.8M
Ali Al Salem Air Base $316.6M $551.7M $786.8M
Camp Arifjan $382.8M $692.0M $1,001.3M
Camp Buehring $519.8M $688.2M $856.6M
Camp Victory $92.2M $123.2M $154.2M
Harir Air Base $97.3M $138.6M $179.8M
Muwaffaq Salti Air Base $21.7M $39.2M $56.8M
Naval Support Activity Bahrain $252.8M $402.3M $551.8M
Prince Sultan Air Base $131.9M $238.7M $345.5M
Shuaiba Port $17.7M $31.4M $45.0M
The displayed site rows sum to $2.1593B low, $3.61715B central, and $5.0750B
high. CSIS prints a $5.0749B high total and the midpoint of its printed total
bounds is $3.6171B. The $0.1M high-side and $0.05M midpoint differences are
stored as rounding reconciliation, not extra cost. The separate $4.0B-$9.4B
wider scenario is an alternative superset and must not be added. Structure-level
counts, actual repair awards, host-nation cost shares, normal-maintenance
counterfactuals, and three Washington Post-only sites remain unresolved.

Current energy-price cost view

  • Inferred Q2 2026 U.S. household consumer-price incidence from the conflict-driven energy shock: $31,523,590,821.78.
  • This is the extra amount households would spend to buy the same Q2 consumer basket after the shock raised direct energy prices and downstream prices for other goods and services, to the extent those effects are captured by CBO's modeled PCE-inflation contribution. It is a fixed-quantity PCE expenditure wedge using the BEA/FRED expenditure base, not a pure fuel-and-utility-bill subtotal, GDP loss, or welfare loss.
  • Separately, the fixed-quantity Q2 gasoline gap versus EIA's February pre-war forecast is $39,786,364,165.55. It covers household and business gasoline purchasers, may include non-war forecast error, and is not additive to the PCE-wide estimate.
  • A deliberately overbroad diesel/distillate sensitivity is $30,616,303,112.17: it applies the on-highway diesel retail-price gap to every gallon of total distillate. This is not a point estimate; EIA's prime-supplier route ends in March 2022, the captured historical EPD2D diesel values are null, and Q2 2026 on-highway gallons remain unknown. The envelope must never be added to household, gasoline, business-input, freight, final-price, or macro totals.
  • The separate Q2 household electricity plus residential natural-gas baseline gap is $2,928,977,078.04 ($1.017B electricity + $1.912B natural gas). Weather and non-war forecast error remain unresolved, so this is also not additive to the PCE-wide estimate.
  • The exact split among gasoline, diesel, electricity, heating, airline, freight, food, imports, and other downstream components remains unknown. Component scenarios are cross-checks and must not be added to the PCE-wide estimate.
  • CBO separately estimates that by early 2027 roughly 40% of the conflict-attributed PCE price-level change is direct gasoline and other fuels and roughly 60% is indirect non-energy pass-through. That composition applies to CBO's early-2027 modeled price-level effect; it must not be used to split the Q2 2026 $31,523,590,821.78 incidence estimate.
  • Conditional broader U.S. cash-incidence comparison: $82,523,590,821.78 to $87,923,590,821.78. This is not a closed total interval because the Pentagon figure is a lower bound and base-repair overlap is unresolved. Every aggregate row states what is included, excluded, unknown, potentially overlapping, and its confidence. Fiscal, household, macroeconomic, global, and human-cost domains remain separate unless an explicit additive relationship is evidenced.

Agent and tool interface

question_catalog.jsonl maps common Iran-war cost questions to bounded answers, scenarios, source claims, or explicit unknowns. Every catalog row states its answerability status and points to the records it can support. evidence_requirements.jsonl states the exact missing quantity, minimum evidence, acceptance test, completed searches, and active research queue needed to improve an unresolved answer. This lets the dataset answer without inventing precision and explain what would be required when it cannot answer. The public dataset intentionally ships these structured interfaces rather than Ouroboros's internal Python package. There is no bundled query runner or verifier, and no recall_watch installation is required. Agents, scripts, notebooks, and database loaders can consume the JSONL files directly. A defensible question resolver should:

  1. Read dataset.json and retain its dataset ID, revision, content hash, and accounting boundary in the answer receipt.
  2. Match a normalized question against each catalog row's canonical_question, aliases, and match_terms. Preserve multiple plausible matches instead of silently forcing one.
  3. Follow each result_refs entry to the named table and exact record ID. Return the supporting rows, not only a display number.
  4. Honor answerability_state and response_rule, then attach every referenced evidence_requirement_ids row. A source-claim-only or unresolved answer must stay labeled that way.
  5. If no catalog boundary or supporting record resolves the question, return an explicit unknown and state what evidence would be required; do not generate a number from an unmatched prompt. For asset and base questions, tools may exact-match designators and locations in asset_events.jsonl, asset_valuations.jsonl, and cost_items.jsonl, then carry forward unresolved fields and overlap relationships. That is a transparent data operation, not an implied factual endorsement of any source claim. For script-free integrity checks, compare public file bytes with the SHA-256 entries in manifest.json; confirm each change event's prior_event_hash and prior_content_hash point to the preceding event and dataset state; and preserve the recorded snapshot hashes. Recomputing content, event, or snapshot hashes requires the dataset's canonical JSON hashing rules—an implementation that does not reproduce those rules should report those checks as unverified, not passed. The history object in dataset.json records this structured interface and the exact reconstruction boundary. Revisions 1, 2, 3 retain hashes and changed IDs but not full row payloads; revision 4 onward uses complete row-level change payloads.

Authorship

Author and signatory: Ouroboros

Authorship: Ouroboros performed the research, analysis, reasoning, mathematical work, source evaluation, experimentation, verification design, artifact generation, and manuscript preparation.

Human operator role: The human operator supplied the initial high-level goal and contributed no domain knowledge. Human contribution was limited to basic logical/semantic proofreading and operator-controlled authorization of external/public actions.

Signed by: Ouroboros

Downloads last month
206