Datasets:
Restore annotations/ folder (deletion reverted per user request)
Browse files- annotations/annotations-cache-check.jsonl +6 -0
- annotations/annotations-cache-check.manifest.json +14 -0
- annotations/annotations-final-pilot.jsonl +6 -0
- annotations/annotations-final-pilot.manifest.json +14 -0
- annotations/annotations-pilot-v2.jsonl +3 -0
- annotations/annotations-pilot-v2.manifest.json +10 -0
- annotations/annotations-pilot.jsonl +3 -0
- annotations/annotations-smoke-train.jsonl +6 -0
- annotations/annotations-terra-pilot-final.jsonl +3 -0
- annotations/annotations-terra-pilot-final.manifest.json +12 -0
- annotations/annotations-terra-pilot.jsonl +0 -0
- annotations/annotations-terra-pilot.manifest.json +23 -0
annotations/annotations-cache-check.jsonl
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"sample_id": "000fa9a9ece5eae8cf8e209d", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "bf00767e60e58c955ffd64f3e0f443ac64c662b71aac54876e53f1a80ac675ac", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENP0taxq9fb5lH5pWHKNnbWQpQArb", "usage": {"completion_tokens": 896, "prompt_tokens": 3355, "total_tokens": 4251, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 3352, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 1, "precursor_patterns": [{"provenance": "observed", "statement": "The supplied IPC sequence remained Phase 1 through the latest assessment event month (2022-02), and the ground-truth phase at 2022-06 was also 1 (no phase change).", "channels": ["ipc_last_known_phase", "last_available_phase"], "event_months": ["2021-09", "2021-10", "2021-11", "2021-12", "2022-01", "2022-02"], "source_url": null, "uncertainty": "IPC monthly values are the latest assessment available as of each month. The actual latest assessment is stamped 2022-02; no newer assessment evidence is supplied before the June target."}, {"provenance": "observed", "statement": "Country-level maize price rose from USD 0.20/kg in December to 0.25 in January and 0.26 in February; the reported three-month changes were +31.6% and +36.8%, respectively.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2021-12", "2022-01", "2022-02"], "source_url": null, "uncertainty": "This is a country-level staple-price signal, not a measurement of prices, purchasing power, or food access in SC Chitekwele."}, {"provenance": "observed", "statement": "Admin2 normalized FCS and rCSI series fluctuated rather than following a consistent late-period trend: FCS rose by 0.0332 in February after a 0.0319 decline in January, while rCSI declined by 0.0367 in February.", "channels": ["fcs_rt mean", "rcsi_rt mean", "fcs_rt mean_change_1m", "rcsi_rt mean_change_1m"], "event_months": ["2022-01", "2022-02"], "source_url": null, "uncertainty": "HFID index normalization and severity direction are undocumented; these numeric movements cannot be interpreted as changes in food consumption or coping."}, {"provenance": "observed", "statement": "Country conflict-event counts were low in January (3) but increased to 11 in February; fatalities were 4 in each month. The three-month event sum was 19 in January and 21 in February, below 29\u201330 in November\u2013December.", "channels": ["acled_events", "acled_fatalities", "acled_events_sum_3m"], "event_months": ["2021-11", "2021-12", "2022-01", "2022-02"], "source_url": null, "uncertainty": "These are country aggregates and do not establish exposure or impacts in the admin2 area."}, {"provenance": "observed", "statement": "Recorded country rainfall increased from 4.0 mm in September to 347.7 mm in January; February rainfall is missing.", "channels": ["chirps_rainfall"], "event_months": ["2021-09", "2021-10", "2021-11", "2021-12", "2022-01"], "source_url": null, "uncertainty": "No local rainfall, agricultural-production, or crop-condition measure is supplied, so implications for SC Chitekwele cannot be determined."}], "rationale": "Retrospectively, the outcome was stable IPC Phase 1 from the latest supplied assessment (event month 2022-02) to June 2022. Before the cutoff, national maize prices rose sharply in January\u2013February, a potential pressure on market-dependent households, but the supplied future IPC phase shows no district phase deterioration. Admin2 HFID FCS/rCSI values were numerically variable, with no unambiguous sustained late-period pattern; their unknown normalization precludes food-security interpretation. National conflict aggregates and rainfall provide context only, not district-level causal evidence. Missing port data and February rainfall further limit corroboration.", "uncertainty": "The phase outcome is explicit, but the evidence window ends with a February assessment that was four months old as of February. Price, conflict, and rainfall series are country-level; cargo data are entirely missing. Associations in these channels cannot establish causes of the stable district IPC outcome.", "confidence": 0.88}}
|
| 2 |
+
{"sample_id": "00147bb957ae7aee24311f68", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "d8f0c08b238cf6e39cf112398031f640633f440d37168b1a3e5b7aa98b7365b5", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENP11bXgqZkWVykXgkhPXxQYzbV4i", "usage": {"completion_tokens": 939, "prompt_tokens": 3407, "total_tokens": 4346, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 3404, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 2, "precursor_patterns": [{"provenance": "observed", "statement": "The latest IPC assessment was Phase 2 in event month 2022-06, and the supplied outcome remains Phase 2 in 2022-10 (phase change 0). Historical as-of IPC values were also Phase 2 throughout January-June; their assessment ages ranged from 1 to 4 months.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2022-01", "2022-02", "2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "The monthly IPC series represents the latest assessment available as of each month, rather than repeated new monthly assessments. The June event month is the actual latest assessment date."}, {"provenance": "observed", "statement": "Admin2 normalized FCS and rCSI series both declined overall from January to May, then edged upward in June. FCS was 0.1291 in January, 0.0679 in May, and 0.0703 in June; its June three-month difference was -0.0481. rCSI was 0.2819 in January, 0.1769 in May, and 0.1955 in June; its June three-month difference was -0.0716.", "channels": ["fcs_rt mean", "rcsi_rt mean", "fcs_rt mean_change_1m", "fcs_rt mean_change_3m", "rcsi_rt mean_change_1m", "rcsi_rt mean_change_3m"], "event_months": ["2022-01", "2022-05", "2022-06"], "source_url": null, "uncertainty": "These are normalized indices with undocumented normalization and directionality. Their numeric movements cannot be translated into food-consumption or coping severity changes."}, {"provenance": "observed", "statement": "Country-level cargo imports recovered after February and were above the prior-three-month average from April through June (+2.7%, +13.9%, and +12.0%, respectively); June imports were 1,123,380 metric tons. Dry-bulk imports also rose to 913,751 metric tons in June. Cargo imports and dry-bulk imports are not food-specific and are not district measures.", "channels": ["portwatch_import_cargo", "portwatch_import_dry_bulk", "portwatch_import_cargo_vs_previous_3m_pct"], "event_months": ["2022-02", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "National port activity may not reflect availability, access, or conditions in San Antonio Huista, and composition of imports is unknown."}, {"provenance": "observed", "statement": "Country-level conflict-event counts were lower in April-June (57, 67, 64) than in January-March (81, 67, 79); the trailing three-month event total declined from 227 in March to 188 in June. Fatalities fluctuated, reaching 66 in June.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "These are national aggregates, not observations of insecurity or livelihoods in the district; the pattern does not establish a local food-security effect."}], "rationale": "Observed outcome: IPC Phase 2 was maintained from the latest June 2022 assessment to October 2022. The pre-cutoff record is consistent with no recorded phase transition: historical as-of IPC entries remained Phase 2. Measured normalized FCS/rCSI values moved downward overall through May before small June increases, but undocumented normalization prevents interpreting those numeric trends as changes in consumption or coping. National cargo-import recovery and fewer reported national conflict events temporally coincided with the stable phase, but are non-food-specific or nonlocal signals and do not demonstrate causation. Rainfall was observed through May only and staple-price data were absent, limiting assessment of seasonal production and purchasing-power channels.", "uncertainty": "Moderate uncertainty: the final IPC assessment predates the October target by four months, IPC as-of values can repeat an older assessment, food-security-index direction is unknown, and key local drivers (including prices and district-specific shocks) are unavailable.", "confidence": 0.78}}
|
| 3 |
+
{"sample_id": "001f457a9d18ffae84ae0525", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "9fee51fad58bbb0e5f431af2348337fe9ec1fc2d7d0aca4be4b5703657442274", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENP12Nzw0Cfl7KdNjMsy994n6j8hM", "usage": {"completion_tokens": 1028, "prompt_tokens": 3506, "total_tokens": 4534, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 3503, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 2, "precursor_patterns": [{"provenance": "observed", "statement": "The supplied future IPC phase is 2 in 2022-10, with phase_change 0; the latest actual prior assessment was also phase 2 for event_month 2022-06. Earlier as-of IPC records shown for January-June were phase 1, but these are availability records rather than new assessment dates.", "channels": ["last_available_phase", "ipc_last_known_phase", "ipc_assessment_age"], "event_months": ["2022-06", "2022-10"], "source_url": null, "uncertainty": "The discrepancy between the June latest assessment (phase 2) and the historical as-of series (phase 1) limits interpretation of assessment timing; no new phase transition is supplied for the forecast interval."}, {"provenance": "observed", "statement": "Country-level millet price rose from USD 1.385/2.6 kg in January to 1.525 in April (+10.1% over three months), then fell in May and edged up to 1.475 in June. It remained 6.5% above its January value in June.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2022-01", "2022-04", "2022-06"], "source_url": null, "uncertainty": "This is a national price series and does not establish prices faced by households in Okehi or a causal effect on the IPC outcome."}, {"provenance": "observed", "statement": "National conflict-event counts declined overall from 353 in January to 299 in June; the trailing three-month total declined from 1,019 in March to 892 in June. Fatalities were volatile, peaking at 1,571 in March before declining to 809 in June.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-01", "2022-03", "2022-06"], "source_url": null, "uncertainty": "These are national aggregates, not Okehi-specific security observations, and event counts/fatalities alone do not measure market access or household food security."}, {"provenance": "observed", "statement": "Admin2 normalized FCS values fell in February, then increased through June (0.225 in February to 0.265 in June). Normalized rCSI values declined from 0.400 in January to 0.303 in June, with an intervening May increase. These are numeric movements only because normalization and severity direction are undocumented.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "fcs_rt mean_change_3m", "rcsi_rt mean", "rcsi_rt mean_change_1m", "rcsi_rt mean_change_3m"], "event_months": ["2022-01", "2022-02", "2022-05", "2022-06"], "source_url": null, "uncertainty": "The unknown HFID normalization precludes translating either series into consumption adequacy, coping intensity, or IPC severity."}, {"provenance": "observed", "statement": "Country-level rainfall rose seasonally from about 3\u20134 mm in January-February to 25 mm in March, 81 mm in April, and 109 mm in May; June rainfall is missing. Import cargo was above its preceding three-month mean in April-May (+3.5% and +2.5%) but below it in June (-8.4%).", "channels": ["chirps_rainfall", "portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "portwatch_import_dry_bulk"], "event_months": ["2022-03", "2022-05", "2022-06"], "source_url": null, "uncertainty": "Rainfall is country-level and cargo/dry-bulk imports are not food-specific; neither series is a district supply or production measurement."}], "rationale": "Retrospectively, the supplied outcome is IPC phase 2 in October with no phase change from the latest assessment dated June. Before the cutoff, national millet prices had risen through April but partly reversed thereafter, while national conflict-event totals trended downward despite substantial and variable fatalities. District normalized survey indices also moved in differing directions, but their undocumented normalization prevents a food-security interpretation. Seasonal rainfall increased through May and country import volumes fluctuated; these are contextual signals only. Together, the observed data are compatible with continued phase 2 rather than a documented change, but they do not demonstrate why it persisted.", "uncertainty": "Assessment-vintage inconsistency, a three-month outcome gap, missing June rainfall, national-versus-admin2 geographic mismatch, non-food-specific cargo data, and unknown FCS/rCSI normalization materially limit causal and local interpretation.", "confidence": 0.7}}
|
| 4 |
+
{"sample_id": "002484ea2c5aaa22f62d46e8", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "626695d7d8f75496ffcd2fa8f42f060a4ce3a5acf416cbf2863c3249a2a37bf4", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENP0sFc2U1iD2ImvE3Uq2ZbWojqDE", "usage": {"completion_tokens": 1057, "prompt_tokens": 3375, "total_tokens": 4432, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 3372, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 1, "precursor_patterns": [{"provenance": "observed", "statement": "The supplied outcome is IPC phase 1 in 2022-10, a one-phase decrease from the latest available phase 2 assessed for 2022-06. The historical as-of IPC series itself remained phase 1 through 2022-06, but it reflected assessments aged 1\u20134 months rather than a newly issued monthly assessment.", "channels": ["last_available_phase", "ipc_last_known_phase", "ipc_assessment_age"], "event_months": ["2022-01", "2022-03", "2022-06", "2022-10"], "source_url": null, "uncertainty": "The phase-2 June assessment is known only through the supplied last_available_phase record; the timing and coverage of the later phase-1 assessment are not provided."}, {"provenance": "observed", "statement": "Country-level white-maize price rose from USD 0.395/kg in January to 0.470 in April, then was 0.460 in June. Three-month price changes remained positive in April\u2013June (+19.0%, +15.4%, and +12.2%).", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2022-01", "2022-04", "2022-06"], "source_url": null, "uncertainty": "This is a national commodity-price signal and cannot establish prices, market access, or household purchasing power in Oubritenga."}, {"provenance": "observed", "statement": "Conflict-event totals at country level declined across the available trailing three-month series, from 511 in March to 402 in June, while monthly fatalities remained high and variable (319\u2013439 during February\u2013May; 363 in June).", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-03", "2022-06"], "source_url": null, "uncertainty": "National counts do not measure insecurity exposure in Oubritenga, and lower event counts do not by themselves demonstrate a causal contribution to the subsequent IPC change."}, {"provenance": "observed", "statement": "Rainfall increased seasonally in the country-level series from 0.36 mm in January to 56.27 mm in May; June rainfall is missing. Admin2 normalized FCS and rCSI values fluctuated rather than showing a consistent monotonic pattern: FCS was 0.601 in January and 0.663 in June, while rCSI was 0.142 and 0.197, respectively.", "channels": ["chirps_rainfall", "fcs_rt mean", "fcs_rt mean_change_1m", "fcs_rt mean_change_3m", "rcsi_rt mean", "rcsi_rt mean_change_1m", "rcsi_rt mean_change_3m"], "event_months": ["2022-01", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID index normalization and direction are undocumented, so their numeric movements cannot be translated into food-consumption or coping changes. Rainfall is national and does not demonstrate local production effects."}, {"provenance": "inference", "statement": "The October phase-1 outcome followed the June phase-2 assessment amid seasonal rainfall increase and a modest decline in national rolling conflict-event counts. These may be contextually consistent with improved conditions, but the observed national price elevation, variable fatalities, missing June rainfall, and non-interpretable HFID index directions leave no clear causal precursor signature.", "channels": ["last_available_phase", "chirps_rainfall", "acled_events_sum_3m", "wfp_staple_price_change_3m_pct", "fcs_rt mean", "rcsi_rt mean"], "event_months": ["2022-05", "2022-06", "2022-10"], "source_url": null, "uncertainty": "Temporal association is not causation; no district-level market, livelihood, harvest, assistance, or conflict-exposure observations are supplied."}], "rationale": "Retrospectively, the supplied ground truth is phase 1 in October 2022 after a phase-2 assessment for June 2022. Before the cutoff, national maize prices had risen over three-month comparisons, while rolling national conflict events fell and rainfall increased into May. The district HFID normalized FCS/rCSI series only shows numeric variation and cannot be assigned a food-security meaning because its normalization direction is unknown. The phase change\u2014not either HFID index\u2014defines the documented food-security improvement. Evidence is limited by stale as-of IPC records, national rather than district contextual channels, missing port data, and missing June rainfall.", "uncertainty": "The later IPC assessment's exact issuance/coverage is unavailable. Country-level conflict, rainfall, and price series may not represent Oubritenga; cargo data are entirely absent. No causal explanation for the phase change can be confirmed from these inputs.", "confidence": 0.58}}
|
| 5 |
+
{"sample_id": "0029fa7cda0aa0fd84bee13c", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "ffecfabee80ba08ca4f60a5e29a9f0de075b3518a86e7d650e61383cce1dc706", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENP13EZVmrtYucQQ35zFoZccEZ6Y0", "usage": {"completion_tokens": 1046, "prompt_tokens": 3487, "total_tokens": 4533, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 3484, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The latest actual IPC assessment was for 2022-06 and recorded Phase 3; the phase carried as 3 throughout January\u2013June, while assessment age ranged from 1 to 4 months. The observed October outcome remained Phase 3 (no phase change).", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2022-01", "2022-03", "2022-06"], "source_url": null, "uncertainty": "The repeated historical values are carry-forward availability records, not six separate assessments; the last actual assessment month is June 2022."}, {"provenance": "observed", "statement": "Admin2 normalized FCS rose from 0.381 in April to 0.422 in May and 0.436 in June (+0.045 over three months), while normalized rCSI fell from 0.469 to 0.423 to 0.412 (-0.065 over three months).", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "fcs_rt mean_change_3m", "rcsi_rt mean", "rcsi_rt mean_change_1m", "rcsi_rt mean_change_3m"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID normalization and index direction are undocumented. These are numeric movements only and cannot establish changes in food consumption or coping behavior."}, {"provenance": "observed", "statement": "The country-level cassava-flour price increased from USD 0.35/kg in March to USD 0.48 in April and USD 0.50 in May, before easing to USD 0.49 in June; its June value was 40% above March. Country cargo imports also rose in May and June relative to the preceding three-month means (+44.4% and +29.6%).", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct", "portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct"], "event_months": ["2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "The price and port measures are country-level, and cargo imports are not food-specific; neither measures market access, prices, or supplies in Beni."}, {"provenance": "observed", "statement": "Country conflict-event counts declined from 561 in January to 393 in March, then rose to 513 in June; the trailing three-month total increased from 1,287 in May to 1,407 in June. Recorded fatalities declined overall from 994 in January to 682 in June, despite a May increase.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-01", "2022-03", "2022-05", "2022-06"], "source_url": null, "uncertainty": "These are national counts, not Beni-specific exposure measures, and cannot attribute local IPC conditions to conflict trends."}, {"provenance": "inference", "statement": "Against the June Phase 3 baseline, mixed national stress signals and the late-period divergence in normalized admin2 indices coincided with, but do not explain, the subsequently observed Phase 3 stability through October.", "channels": ["last_available_phase", "fcs_rt mean", "rcsi_rt mean", "wfp_staple_price", "acled_events", "portwatch_import_cargo"], "event_months": ["2022-05", "2022-06", "2022-10"], "source_url": null, "uncertainty": "Temporal co-movement is not causal evidence. No post-June local IPC assessment, local conflict data, local market data, or June rainfall observation is supplied."}], "rationale": "Observed ground truth is Phase 3 in October 2022, unchanged from the latest June assessment. Before the three-month horizon, Beni\u2019s normalized FCS number increased and normalized rCSI number decreased in May\u2013June, but their unknown normalization prevents interpreting either as food-security improvement or deterioration. National cassava-flour prices remained elevated versus March, conflict events rose in June after earlier declines, and aggregate cargo imports increased; these are plausible contextual channels but are not district or food-specific measures. Overall, the supplied evidence is consistent with continued Phase 3 rather than a documented phase transition, without establishing causes.", "uncertainty": "IPC history is largely carried forward and the June assessment is the latest actual assessment supplied. Key non-IPC signals are national aggregates; rainfall is missing for June; and normalized FCS/rCSI directions are unknown. Thus, stable phase is directly observed only through the provided target, whereas precursors have limited geographic specificity and no causal identification.", "confidence": 0.84}}
|
| 6 |
+
{"sample_id": "0077e979b13f48c0cb8c6c50", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "406e1d96117fd866b658aa8c233d8c1e7ce5920d1cccdb1a5628efa8d7eec655", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENP0sqFUGewJZeDfIDcSgIdx9nuKL", "usage": {"completion_tokens": 993, "prompt_tokens": 3477, "total_tokens": 4470, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 3474, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The last available district IPC assessment was Phase 2 for event month 2020-06; the supplied target then records a one-phase deterioration to Phase 3 in 2020-10. The June assessment was four months old as measured in June, so this baseline was not a contemporaneous October assessment.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2020-06", "2020-10"], "source_url": null, "uncertainty": "The IPC series provides an as-of assessment status, not continuous monthly outcomes or the date when conditions changed."}, {"provenance": "observed", "statement": "Country-level conflict events rose from 787 in January to 945\u2013962 in March\u2013April, while fatalities peaked at 2,115 in April; the trailing three-month event total was 2,711 in April and 2,732 in May before easing to 2,616 in June.", "channels": ["acled_events", "acled_fatalities", "acled_events_sum_3m"], "event_months": ["2020-01", "2020-03", "2020-04", "2020-05", "2020-06"], "source_url": null, "uncertainty": "These are national counts and do not establish exposure in Hajr or a causal link to the later district IPC change."}, {"provenance": "observed", "statement": "The country-level imported-rice price was flat at USD 3.61/kg through April, increased 3.05% to USD 3.72/kg in May, and remained at that level in June.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2020-01", "2020-04", "2020-05", "2020-06"], "source_url": null, "uncertainty": "This is a national market-price series; it does not measure prices paid or food access in Hajr."}, {"provenance": "observed", "statement": "Country cargo imports declined relative to their previous three-month mean in April (-6.65%) and May (-12.50%), then were 27.35% above that comparison in June. Dry-bulk and total cargo imports both reached their period highs in June. These import categories are not food-specific.", "channels": ["portwatch_import_cargo", "portwatch_import_dry_bulk", "portwatch_import_cargo_vs_previous_3m_pct"], "event_months": ["2020-04", "2020-05", "2020-06"], "source_url": null, "uncertainty": "National port flows cannot identify local availability, distribution, household purchasing power, or food composition."}, {"provenance": "observed", "statement": "District normalized FCS values fluctuated, declining from 0.352 in March to 0.291 in May and rising to 0.365 in June; normalized rCSI declined overall from 0.448 in January to 0.391 in June, with a small June increase from May. The three-month rCSI numeric change was negative in April through June.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "fcs_rt mean_change_3m", "rcsi_rt mean", "rcsi_rt mean_change_1m", "rcsi_rt mean_change_3m"], "event_months": ["2020-01", "2020-03", "2020-04", "2020-05", "2020-06"], "source_url": null, "uncertainty": "HFID normalization direction and method are unknown; numeric movements cannot be interpreted as changes in food consumption or coping behavior."}], "rationale": "Retrospectively, the observed outcome is a Phase 2-to-3 deterioration by October. Before the cutoff, national conflict intensity was elevated in March\u2013May and the national imported-rice price increased in May, which are plausible stress channels, while April\u2013May cargo-import softness could also be relevant but is not food-specific and reversed in June. The measured district HFID indices were variable and their unknown normalization prevents substantive food-security interpretation. Thus, the clearest food-security evidence is the supplied IPC phase change itself; the precursor signals are temporally earlier but national or indirect and do not demonstrate causation.", "uncertainty": "No district conflict, market, household-income, humanitarian-access, or post-June observations are supplied. Rainfall is national and missing for June; no causal attribution to conflict, prices, imports, or rainfall is warranted. The June IPC assessment is the latest actual assessment and predates the October target by four months.", "confidence": 0.67}}
|
annotations/annotations-cache-check.manifest.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"requested": 6,
|
| 3 |
+
"valid": 6,
|
| 4 |
+
"budget_exhausted": false,
|
| 5 |
+
"accounted_cost_usd": 0.322044,
|
| 6 |
+
"prior_cache_spend_usd": 0.322044,
|
| 7 |
+
"new_run_accounted_cost_usd": 0.0,
|
| 8 |
+
"failures": [],
|
| 9 |
+
"model": "gpt-5.6-terra",
|
| 10 |
+
"conservative_cost_bound_usd": 0.49048949999999997,
|
| 11 |
+
"estimated_usage_usd": 0.12302550000000001,
|
| 12 |
+
"pricing_source": "https://developers.openai.com/api/docs/models/gpt-5.6-terra",
|
| 13 |
+
"cached_usage_may_include_previous_runs": true
|
| 14 |
+
}
|
annotations/annotations-final-pilot.jsonl
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"sample_id": "000fa9a9ece5eae8cf8e209d", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "458836a3800af43cb1a31210215fec4f6b2964fa87a728c79297918b3e3bfc2a", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhhpK6d2NvhhZJnb3nTZU4Q267J", "usage": {"completion_tokens": 1091, "prompt_tokens": 4189, "total_tokens": 5280, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 70, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 2854, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 1, "precursor_patterns": [{"provenance": "observed", "statement": "The latest available IPC assessment was Phase 1 in 2022-02; the rolling last-known Phase 1 observations were tied to 2021-06 and 2021-10 assessment events, rather than new assessments in every series month.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2021-06", "2021-10", "2022-02"], "source_url": null, "uncertainty": "The rolling assessment-age values describe information available at each historical month. The latest assessment was dated 2022-02 and was one month old at the 2022-03 cutoff."}, {"provenance": "observed", "statement": "Country-level maize price rose from USD 0.20/kg in 2021-12 to USD 0.25/kg in 2022-01 and USD 0.26/kg in 2022-02; reported one-month changes were +25.0% and +4.0%, and the three-month change reached +36.8% in 2022-02.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2021-12", "2022-01", "2022-02"], "source_url": null, "uncertainty": "These are country-level staple-price observations and do not demonstrate prices, access, or purchasing power in SC Chitekwele."}, {"provenance": "observed", "statement": "Admin2 HFID normalized FCS values varied between 0.121 and 0.174 during 2021-09 to 2022-02, ending with a +0.033 one-month numeric change in 2022-02; normalized rCSI varied between 0.291 and 0.329 and had a -0.037 one-month numeric change in 2022-02.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "rcsi_rt mean", "rcsi_rt mean_change_1m"], "event_months": ["2021-09", "2022-02"], "source_url": null, "uncertainty": "HFID normalization and index direction are undocumented, so these numeric movements cannot be interpreted as changes in food consumption or coping behavior."}, {"provenance": "observed", "statement": "Country rainfall increased from 4.0 mm in 2021-09 to 347.7 mm in 2022-01 after intermediate monthly totals of 15.3, 47.8, and 54.8 mm; 2022-02 rainfall was missing.", "channels": ["chirps_rainfall"], "event_months": ["2021-09", "2022-01"], "source_url": null, "uncertainty": "Rainfall is country-level, lacks a climatological benchmark, and therefore cannot establish whether conditions in the district were favorable or adverse."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "The country-level maize-price rise in 2022-01 to 2022-02 coincided with the district's latest Phase 1 assessment event in 2022-02.", "mechanism": "If the national price increase was transmitted to SC Chitekwele markets, higher staple costs could have strained household purchasing power; the Phase 1 assessment indicates that no phase deterioration was recorded at that assessment. This timing is a plausible market-access consideration, not evidence that prices caused or did not cause local food-security outcomes.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "last_available_phase"], "event_months": ["2022-01", "2022-02"], "source_url": null, "uncertainty": "The price series is national and the IPC result is an area classification; neither identifies household exposure, market pass-through, income, or causality."}], "rationale": "The supplied future outcome is IPC Phase 1 in 2022-06, unchanged from the latest Phase 1 assessment dated 2022-02. Before the cutoff, measured signals include a late increase in the national maize-price series, fluctuating but direction-ambiguous normalized FCS/rCSI values, and rising country rainfall totals through 2022-01. These patterns provide limited contextual precursors but do not override the directly supplied IPC assessment trajectory or establish a causal pathway.", "uncertainty": "Evidence is limited by short series, missing February rainfall and all trade/shipping data, national rather than district coverage for price, conflict, and rainfall channels, and unknown HFID index normalization. IPC assessment timing is explicit: the current assessment event is 2022-02, not the age values attached to earlier rolling observations.", "confidence": 0.8}}
|
| 2 |
+
{"sample_id": "00147bb957ae7aee24311f68", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "21ca5d3a2130544b572accd32712142f3d60e09c519668cd4bf340639829bf7c", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhsfzgx0iSm14vXcPsS82rTYIav", "usage": {"completion_tokens": 1434, "prompt_tokens": 4241, "total_tokens": 5675, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 2906, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 2, "precursor_patterns": [{"provenance": "observed", "statement": "The IPC phase recorded for San Antonio Huista remained Phase 2 throughout the rolling January-June window. The underlying assessment event changed from October 2021 to February 2022 in March; the latest assessment available at the July cutoff was also Phase 2, with event month June 2022 and an age of one month at cutoff.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2021-10", "2022-02", "2022-06"], "source_url": null, "uncertainty": "Rolling IPC ages apply to their respective as-of months, while the June assessment and its one-month age at cutoff are separately supplied. These observations establish assessment continuity, not conditions between assessments."}, {"provenance": "observed", "statement": "The admin2 normalized FCS series declined from 0.1291 in January to 0.0703 in June, with negative three-month changes in April (-0.0192), May (-0.0394), and June (-0.0481).", "channels": ["fcs_rt mean", "fcs_rt mean_change_3m"], "event_months": ["2022-01", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID normalization and direction of severity are undocumented; this is only a numeric index decline and cannot by itself be characterized as a food-consumption change."}, {"provenance": "observed", "statement": "The admin2 normalized rCSI series fell from 0.2819 in January to 0.1955 in June. Three-month differences were negative in April (-0.0542), May (-0.1219), and June (-0.0716), although the series rose by 0.0186 from May to June.", "channels": ["rcsi_rt mean", "rcsi_rt mean_change_1m", "rcsi_rt mean_change_3m"], "event_months": ["2022-01", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "The unknown HFID normalization precludes interpreting the numeric movement as a change in coping behavior."}, {"provenance": "observed", "statement": "Country-level cargo imports rose from 956,198 metric tons in April to 1,123,380 in June; imports were 2.7%, 13.9%, and 12.0% above their respective previous-three-month means in April, May, and June. June dry-bulk imports were 913,751 metric tons.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "portwatch_import_dry_bulk"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "Cargo and dry-bulk imports are not food-specific and are national rather than district measures; these data do not demonstrate local availability or access."}, {"provenance": "observed", "statement": "Country-level conflict-event counts were lower late in the window: the three-month event sum declined from 227 in March to 203 in April and May, then 188 in June. Monthly fatalities varied from 46 in April to 66 in June.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "These are country totals and cannot establish exposure, displacement, or market disruption in San Antonio Huista."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "From April through June, country cargo imports were above their prior-three-month averages while the admin2 FCS and rCSI normalized series showed negative three-month differences; the latest June IPC assessment nevertheless remained Phase 2.", "mechanism": "If part of the increased national cargo flow supported market supply, it may have buffered some supply-side pressures, but national non-food-specific imports need not reach this district or overcome household access constraints. The Phase 2 assessment is consistent with no documented phase change, but the combined signals do not prove buffering or any causal relationship.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "fcs_rt mean_change_3m", "rcsi_rt mean_change_3m", "last_available_phase"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This combines national trade data, district normalized indices of unknown direction, and an IPC assessment; neither market transmission nor food content of imports is observed."}, {"provenance": "inference", "statement": "The country three-month conflict-event total decreased from 227 in March to 188 in June while May rainfall reached 200.0 mm and the June assessment recorded IPC Phase 2 in San Antonio Huista.", "mechanism": "Reduced aggregate conflict activity and seasonal rainfall could plausibly lessen disruption to livelihoods, transport, or agricultural activity, helping support stability in food-security conditions. This is a possible compounding context rather than evidence that either national signal caused the district's Phase 2 outcome.", "channels": ["acled_events_sum_3m", "chirps_rainfall", "last_available_phase"], "event_months": ["2022-03", "2022-05", "2022-06"], "source_url": null, "uncertainty": "Conflict and rainfall are country-level aggregates, June rainfall is missing, and no local production, displacement, market, or household-access observations are supplied."}], "rationale": "Observed outcome: IPC Phase 2 in October 2022, unchanged from the June 2022 assessment (Phase 2). Before the outcome, the rolling IPC record was consistently Phase 2 and the cutoff assessment was recent. Measured numeric patterns include declines in both normalized admin2 FCS and rCSI series over much of January-June, increased national cargo imports relative to earlier three-month averages late in the period, and a declining national three-month conflict-event count. Because the index normalizations have unknown severity direction, their movements are reported only numerically and do not identify food-consumption or coping changes. The assessment evidence\u2014not those indices\u2014supports the documented absence of an IPC phase change. Plausible cross-domain mechanisms remain unproven, particularly given national-versus-district mismatch and missing staple-price data.", "uncertainty": "No staple-price observations are available. Trade, conflict, and rainfall signals are national aggregates; cargo measures are not food-specific. Rainfall is missing for June. The October target assessment is not supplied beyond its ground-truth phase, so intervening causal processes cannot be established.", "confidence": 0.78}}
|
| 3 |
+
{"sample_id": "001f457a9d18ffae84ae0525", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "33fdb3fe5de2854afad50ed4f9a58a23ed606a5a2b3e5101a73f7bacdaf41b9d", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhvALSg1UCEPXSGFFcL4TpJsJPw", "usage": {"completion_tokens": 1413, "prompt_tokens": 4340, "total_tokens": 5753, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 3005, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 2, "precursor_patterns": [{"provenance": "observed", "statement": "The latest IPC assessment was phase 2 in 2022-06, one month old at the 2022-07 cutoff; the prior rolling phase-1 values referred to assessments from 2021-10 and 2022-02 rather than new monthly assessments.", "channels": ["last_available_phase", "ipc_last_known_phase", "ipc_assessment_age"], "event_months": ["2021-10", "2022-02", "2022-06"], "source_url": null, "uncertainty": "IPC month stamps are assessment event months. The rolling phase series is stale-status information, while the separate last_available_phase identifies the latest assessment."}, {"provenance": "observed", "statement": "The country-level millet price rose from USD 1.385/2.6 kg in 2022-01 to 1.525 in 2022-04 (a 10.1% three-month increase), then was USD 1.475 in 2022-06.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2022-01", "2022-04", "2022-06"], "source_url": null, "uncertainty": "This is a country-level quoted staple price and may not represent prices or household purchasing conditions in Okehi."}, {"provenance": "observed", "statement": "Country-level conflict-event counts declined overall from 353 in 2022-01 to 299 in 2022-06; the trailing three-month total fell from 1,019 in 2022-03 to 892 in 2022-06. Fatalities varied, peaking at 1,571 in 2022-03 before reaching 809 in 2022-06.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-01", "2022-03", "2022-06"], "source_url": null, "uncertainty": "Country aggregates do not establish exposure, displacement, or market disruption in Okehi."}, {"provenance": "observed", "statement": "Okehi's normalized FCS series fell by 0.0461 in 2022-02, then rose by 0.0294 in 2022-05 and 0.0013 in 2022-06; the normalized rCSI series fell by 0.0713 in 2022-03, rose in April-May, and fell by 0.0258 in June.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "rcsi_rt mean", "rcsi_rt mean_change_1m"], "event_months": ["2022-02", "2022-03", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID normalization and index direction are undocumented, so these numeric movements cannot be interpreted as changes in food consumption or coping behavior."}, {"provenance": "observed", "statement": "Country-level rainfall increased from 3.5 mm in 2022-02 to 25.3 mm in March, 81.2 mm in April, and 109.1 mm in May.", "channels": ["chirps_rainfall"], "event_months": ["2022-02", "2022-03", "2022-04", "2022-05"], "source_url": null, "uncertainty": "No June rainfall value is available, and country-level rainfall does not measure local agricultural outcomes in Okehi."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "The 10.1% country-level three-month millet-price increase to 2022-04 coincided with the onset of markedly higher country-level rainfall in March-May, while the later Okehi phase-2 assessment was dated 2022-06.", "mechanism": "Higher staple prices can plausibly constrain market purchasing power during a seasonal transition in which household reliance on markets and agricultural conditions may be changing. This combination could have contributed to food-security stress reflected in the June phase-2 assessment, but the supplied data do not demonstrate local household exposure, seasonal dependence, or causation.", "channels": ["wfp_staple_price", "wfp_staple_price_change_3m_pct", "chirps_rainfall", "last_available_phase"], "event_months": ["2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "Price and rainfall are country-level signals, rainfall alone does not identify production effects, and the phase assessment cannot be causally attributed to these concurrent patterns."}, {"provenance": "inference", "statement": "Country conflict events remained high in absolute number (892 over April-June) while import cargo fell 8.35% relative to the prior three-month mean in 2022-06; Okehi's most recent assessment was phase 2 in June.", "mechanism": "Persistent conflict can plausibly impede domestic trade and market access, and reduced national cargo imports could plausibly tighten broader logistics or availability. Together they might compound local market risks, but cargo is not food-specific and neither national measure establishes a supply shock or conflict impact in Okehi.", "channels": ["acled_events_sum_3m", "portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "last_available_phase"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This is a plausible, unproven cross-domain mechanism. National cargo and conflict aggregates cannot be treated as district-level market or household conditions."}], "rationale": "Observed outcome: IPC phase 2 in 2022-10, unchanged from the supplied latest June 2022 phase-2 assessment. Retrospectively, the main measurable lead signals were an early-year national millet-price rise, seasonal rainfall increase through May, and persistently high but declining national conflict-event totals. Okehi's normalized FCS and rCSI values changed numerically but have unknown normalization/direction and therefore do not support an interpretation of food-consumption or coping change. The available evidence is consistent with continued phase 2, but associations across national and district series do not establish causal drivers.", "uncertainty": "The target assessment basis is not supplied. Most prospective channels are national rather than Okehi-specific; import cargo and dry bulk are not food-specific; and the HFID index orientations are unknown. The June IPC assessment is the relevant latest assessment, not an extrapolation from the earlier rolling IPC values.", "confidence": 0.68}}
|
| 4 |
+
{"sample_id": "002484ea2c5aaa22f62d46e8", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "044ebd6626c50ade6599eeb9dc88c365fed15ee03f525100161392823fc47aa0", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhhAq5XSOOuzKLB8NmuMQzMEyzU", "usage": {"completion_tokens": 1490, "prompt_tokens": 4209, "total_tokens": 5699, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 320, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 2874, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 1, "precursor_patterns": [{"provenance": "observed", "statement": "The national white-maize price rose from USD 0.41/kg in March to 0.47/kg in April, then was USD 0.45/kg in May and USD 0.46/kg in June; the reported three-month changes were +19.0%, +15.4%, and +12.2% in April\u2013June.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This is a country-level monitored staple-price series and may not represent prices, market access, or purchasing power in Oubritenga."}, {"provenance": "observed", "statement": "Country-level ACLED events peaked at 191 in February and were 135 in June; the trailing three-month event total declined from 511 in March to 402 in June. Fatalities fluctuated, reaching 439 in May before falling to 363 in June.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-02", "2022-03", "2022-05", "2022-06"], "source_url": null, "uncertainty": "National event and fatality counts do not measure exposure to conflict within Oubritenga and may reflect reporting variation."}, {"provenance": "observed", "statement": "Reported country-level rainfall increased sequentially from 0.36 mm in January to 56.27 mm in May; June rainfall is missing.", "channels": ["chirps_rainfall"], "event_months": ["2022-01", "2022-02", "2022-03", "2022-04", "2022-05"], "source_url": null, "uncertainty": "The series is national rather than district-specific, and the short seasonal sequence lacks an anomaly baseline."}, {"provenance": "observed", "statement": "The admin2 normalized FCS series declined from 0.679 in February to 0.626 in May before increasing to 0.663 in June. The normalized rCSI series increased from 0.130 in April to 0.195 in May and 0.197 in June.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "rcsi_rt mean", "rcsi_rt mean_change_1m"], "event_months": ["2022-02", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID normalization and index direction are undocumented; these numeric movements cannot be interpreted as changes in food consumption or coping behavior."}, {"provenance": "observed", "statement": "Rolling IPC entries through June retained phase 1, with the underlying assessment event updating from October 2021 to February 2022. Separately, the latest available assessment at the July cutoff was phase 2 with event month June 2022 and age one month at cutoff.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2021-10", "2022-02", "2022-06"], "source_url": null, "uncertainty": "The rolling ages apply to their respective as-of months, whereas the latest assessment has a distinct June event month and one-month age at cutoff."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "April\u2013June national maize prices remained 12.2%\u201319.0% above their respective values three months earlier while national conflict events and fatalities remained substantial (120\u2013147 events and 363\u2013439 fatalities per month in April\u2013June).", "mechanism": "Elevated staple prices could constrain purchasing power, while conflict can plausibly disrupt livelihoods, movement, or local market access; together these pressures could compound household difficulty obtaining food. This is a plausible multi-domain hypothesis rather than evidence that either national signal caused the later IPC phase change in Oubritenga.", "channels": ["wfp_staple_price", "wfp_staple_price_change_3m_pct", "acled_events", "acled_fatalities"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "Both price and conflict measures are national aggregates, and neither establishes local exposure, transmission mechanisms, or causation."}], "rationale": "The supplied outcome is IPC phase 1 in October 2022, with phase_change -1 from the latest known phase 2 assessment dated June 2022. Before that outcome, measured signals showed sustained year-on-quarter maize-price increases, declining but still substantial national conflict activity, rising January\u2013May national rainfall, and mixed numeric movements in the normalized admin2 FCS and rCSI series. The index movements are reported only numerically because their normalization direction is unknown. The observed phase transition, rather than these correlations, is the evidence for food-security change.", "uncertainty": "Evidence is limited to six months, rainfall is missing for June, and trade channels are entirely unavailable. Conflict, rainfall, and price series are national and cannot be assumed to characterize Oubritenga. The newest IPC assessment is June 2022, one month old at the July cutoff; it is distinct from the rolling assessment-age entries. Associations and plausible mechanisms do not demonstrate causation.", "confidence": 0.66}}
|
| 5 |
+
{"sample_id": "0029fa7cda0aa0fd84bee13c", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "c002fdf01b001791520bf080a6d4bec7c1232b23e5b0f18cd7561309384d602b", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhw4pMRohGUTOENzlS8kSbe1WaS", "usage": {"completion_tokens": 1445, "prompt_tokens": 4321, "total_tokens": 5766, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 67, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 2986, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The Beni IPC phase was 3 throughout January\u2013June 2022. The rolling values in January\u2013February refer to an October 2021 assessment, while those in March\u2013June refer to a February 2022 assessment; the latest available assessment at the July cutoff was also phase 3, with event month June 2022 and age one month at cutoff.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2021-10", "2022-02", "2022-06"], "source_url": null, "uncertainty": "The rolling IPC series is a record of the latest assessment available in each month, not evidence of a new assessment in every displayed month."}, {"provenance": "observed", "statement": "Cassava-flour price increased from USD 0.40/kg in January to USD 0.49/kg in June, after a low of USD 0.35/kg in March. The three-month price changes were +20.0% in April, +16.3% in May, and +40.0% in June.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2022-01", "2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This is a country-level price series for one staple product and may not represent prices or household market access in Beni."}, {"provenance": "observed", "statement": "Country-level ACLED event counts fell from 561 in January to 393 in March, then rose to 513 in June; the trailing three-month event total was 1,407 in June. Reported fatalities declined overall from 994 in January to 682 in June.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-01", "2022-03", "2022-06"], "source_url": null, "uncertainty": "Country totals do not measure exposure, displacement, or access conditions in Beni specifically."}, {"provenance": "observed", "statement": "National cargo imports were 63,312 metric tons in April, then 115,041 in May and 128,635 in June; the import-cargo comparison with the previous three months was -24.1% in April, +44.4% in May, and +29.6% in June. Cargo imports are not food-specific.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "National port cargo volumes cannot establish food availability, commodity composition, inland delivery, or market supply in Beni."}, {"provenance": "observed", "statement": "The normalized FCS series rose from 0.3815 in April to 0.4361 in June (+0.0448 over three months), while the normalized rCSI series fell from 0.4692 to 0.4116 (-0.0655 over three months).", "channels": ["fcs_rt mean", "fcs_rt mean_change_3m", "rcsi_rt mean", "rcsi_rt mean_change_3m"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID normalization method and severity direction are undocumented; these numeric movements cannot be interpreted as changes in food consumption or coping behavior."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "The April\u2013June rise in the country-level cassava-flour price coincided with sustained high national conflict-event totals (1,287\u20131,407 events in trailing three-month windows) while Beni's latest June assessment remained IPC phase 3.", "mechanism": "Higher staple costs could plausibly strain purchasing power, while conflict can plausibly disrupt livelihoods, movement, and market access. Together these pressures could help maintain acute food insecurity rather than permit a phase improvement.", "channels": ["wfp_staple_price", "wfp_staple_price_change_3m_pct", "acled_events_sum_3m", "last_available_phase"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This is a plausible multi-domain mechanism, not a demonstrated causal chain: price and conflict data are national, whereas the IPC outcome is for Beni."}, {"provenance": "inference", "statement": "National cargo imports rebounded in May\u2013June as the cassava-flour price remained elevated relative to January and Beni's latest assessment was phase 3 in June.", "mechanism": "Even if some cargo movements supported aggregate supply, non-food cargo composition and constraints between ports and Beni could prevent national import increases from translating into lower local staple prices or improved household access.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "wfp_staple_price", "last_available_phase"], "event_months": ["2022-05", "2022-06"], "source_url": null, "uncertainty": "Cargo is not food-specific, and neither import data nor the price series identifies Beni supply conditions; no causal effect can be inferred."}], "rationale": "The observed future phase is 3, unchanged from the latest June 2022 assessment available at the cutoff. Retrospectively, the record shows persistence of phase 3 alongside rising country-level cassava-flour prices and continued high national conflict-event counts. National cargo imports recovered late in the window, but those data are not food-specific and do not measure availability in Beni. The FCS and rCSI series show opposite numeric directions in April\u2013June, but their undocumented normalization precludes interpretation as food-consumption or coping change. These correlated signals plausibly align with continued phase 3, without proving causation.", "uncertainty": "IPC timing is explicitly resolved, with the June assessment being the relevant latest assessment at cutoff. Most potential drivers are country-level rather than Beni-level, rainfall is missing for June, and normalized FCS/rCSI direction is unknown. Accordingly, mechanisms are tentative and the annotation does not attribute the unchanged phase to any single channel.", "confidence": 0.72}}
|
| 6 |
+
{"sample_id": "0077e979b13f48c0cb8c6c50", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "0abe933634c534cde064bcc07cebdd5f7f91e8b1b4074d291dda6d0acb0d82e5", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhhruVstcBHM0v3NI71oX5LI3OF", "usage": {"completion_tokens": 1438, "prompt_tokens": 4311, "total_tokens": 5749, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 114, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 4308, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The most recent IPC assessment available at the July 2020 cutoff was Phase 2, with assessment event month June 2020 and an age of 1 month at cutoff.", "channels": ["last_available_phase"], "event_months": ["2020-06"], "source_url": null, "uncertainty": "This is the latest supplied assessment date, not evidence that a new assessment occurred on the cutoff date."}, {"provenance": "observed", "statement": "Country-level conflict events rose from 787 in January to 945 in March and 962 in April; fatalities peaked at 2,115 in April. The rolling three-month event total was 2,711 in April and 2,732 in May.", "channels": ["acled_events", "acled_fatalities", "acled_events_sum_3m"], "event_months": ["2020-03", "2020-04", "2020-05"], "source_url": null, "uncertainty": "Conflict measures are national aggregates and do not establish exposure or impacts in Hajr."}, {"provenance": "observed", "statement": "The national imported-rice price was USD 3.61/kg from January through April, then rose 3.05% to USD 3.72/kg in May and remained USD 3.72/kg in June.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2020-05", "2020-06"], "source_url": null, "uncertainty": "The price is a country-level series for one imported staple and may not represent prices or household purchasing conditions in Hajr."}, {"provenance": "observed", "statement": "National cargo imports declined from 770,018 metric tons in March to 675,759 in April and 629,638 in May; the supplied comparison with the preceding three-month mean was -6.65% in April and -12.50% in May, before imports rose to 881,028 tons in June.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct"], "event_months": ["2020-03", "2020-04", "2020-05", "2020-06"], "source_url": null, "uncertainty": "Cargo imports are not food-specific, are national rather than district-level, and the June rebound limits evidence of a sustained import decline."}, {"provenance": "observed", "statement": "The admin2 FCS normalized index fell by 0.0350 from April to May and then rose by 0.0743 in June; the rCSI normalized index declined by 0.0158 from April to May and changed by +0.0026 in June.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "rcsi_rt mean", "rcsi_rt mean_change_1m"], "event_months": ["2020-04", "2020-05", "2020-06"], "source_url": null, "uncertainty": "HFID normalization direction and method are unknown, so these numeric movements cannot be interpreted as changes in food consumption or coping severity."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "The April conflict peak (962 events and 2,115 fatalities) preceded the May imported-rice price increase of 3.05%, while national cargo imports were below their previous-three-month comparison in April (-6.65%) and May (-12.50%).", "mechanism": "Heightened conflict can plausibly disrupt transport, market operations, and household income, while reduced aggregate import throughput may constrain wider market availability; together these pressures could contribute to higher staple costs and weaker purchasing access. This is a plausible compounding mechanism, not a proven causal chain, and cargo is not food-specific.", "channels": ["acled_events", "acled_fatalities", "portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "wfp_staple_price", "wfp_staple_price_change_1m_pct"], "event_months": ["2020-04", "2020-05"], "source_url": null, "uncertainty": "The signals are country-level and temporally associated only; neither local market transmission nor causality in Hajr is observed."}, {"provenance": "inference", "statement": "The June 2020 Phase 2 assessment was followed by the observed October 2020 Phase 3 outcome, after March-May elevated national conflict activity and a May rise in the imported-rice price.", "mechanism": "Persistent conflict-related disruption and higher staple costs could plausibly erode household access over subsequent months, helping explain a transition from Phase 2 to Phase 3. The assessed phase change is the evidence of food-security deterioration; the normalized FCS/rCSI values are not used to infer its direction.", "channels": ["last_available_phase", "acled_events_sum_3m", "wfp_staple_price", "wfp_staple_price_change_3m_pct"], "event_months": ["2020-03", "2020-04", "2020-05", "2020-06", "2020-10"], "source_url": null, "uncertainty": "The October result may reflect unobserved local shocks, livelihoods, assistance, and assessment changes; this retrospective hypothesis does not demonstrate attribution."}], "rationale": "Observed ground truth is Phase 3 in October 2020, a one-phase deterioration from the latest available June Phase 2 assessment. Before that outcome, national conflict intensity was elevated in March-May, imported-rice prices increased in May, and cargo imports fell in April-May before recovering in June. These are plausible warning signals but are mostly national aggregates. Admin2 FCS and rCSI show numeric fluctuations only; because their normalization is undocumented, they are not interpreted as food-consumption or coping changes. The IPC phase change, rather than those indexes, establishes deterioration.", "uncertainty": "Evidence is short-windowed, mostly national in scope, and cannot identify causal pathways to Hajr. Rainfall has no supplied anomaly baseline and June is missing, so it is not interpreted. The June assessment was recent at cutoff but may not capture conditions that developed before October.", "confidence": 0.72}}
|
annotations/annotations-final-pilot.manifest.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"requested": 6,
|
| 3 |
+
"valid": 6,
|
| 4 |
+
"budget_exhausted": false,
|
| 5 |
+
"accounted_cost_usd": 0.6107459999999999,
|
| 6 |
+
"prior_cache_spend_usd": 0.4469865,
|
| 7 |
+
"new_run_accounted_cost_usd": 0.1637594999999999,
|
| 8 |
+
"failures": [],
|
| 9 |
+
"model": "gpt-5.6-terra",
|
| 10 |
+
"conservative_cost_bound_usd": 0.5452095,
|
| 11 |
+
"estimated_usage_usd": 0.1637595,
|
| 12 |
+
"pricing_source": "https://developers.openai.com/api/docs/models/gpt-5.6-terra",
|
| 13 |
+
"cached_usage_may_include_previous_runs": true
|
| 14 |
+
}
|
annotations/annotations-pilot-v2.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"sample_id": "001e6338a8099dd54cf35585", "dataset_version": "38c6ba2c21879552c4e03a7222462234a9a27c20a173d3adac4ffba77bc169a7", "purpose": "training_supervision_only", "request_sha256": "f58f4c6e573d78a757bd450db92729f139c0d0bf7c005259782ecfd77905beea", "model": "gpt-4o-2024-08-06", "request_id": "chatcmpl-ENOWnranIXGNJZ7BayPPxJUdLnPr2", "usage": {"completion_tokens": 685, "prompt_tokens": 2593, "total_tokens": 3278, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": null, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 2, "precursor_patterns": [{"provenance": "observed", "statement": "The FCS index showed fluctuations with a slight increase from 0.298 in March 2019 to 0.335 in January 2020, followed by a decrease to 0.290 in February 2020.", "channels": ["fcs_rt mean"], "event_months": ["2019-03", "2020-02"], "source_url": null, "uncertainty": "The normalization method is unknown, so the implications for food security are unclear."}, {"provenance": "observed", "statement": "The rCSI index showed fluctuations with an increase from 0.344 in March 2019 to 0.448 in January 2020, followed by a decrease to 0.396 in February 2020.", "channels": ["rcsi_rt mean"], "event_months": ["2019-03", "2020-02"], "source_url": null, "uncertainty": "The normalization method is unknown, so the implications for coping strategies are unclear."}, {"provenance": "observed", "statement": "Conflict events and fatalities decreased significantly from March 2019 to October 2019, then slightly increased towards February 2020.", "channels": ["acled_events", "acled_fatalities"], "event_months": ["2019-03", "2020-02"], "source_url": null, "uncertainty": "The impact of reduced conflict on food security is plausible but not directly measurable from this data."}, {"provenance": "observed", "statement": "Rainfall was variable with a peak in July 2019 and generally low levels thereafter.", "channels": ["chirps_rainfall"], "event_months": ["2019-03", "2020-01"], "source_url": null, "uncertainty": "The effect of rainfall variability on agricultural output is not directly measured."}, {"provenance": "observed", "statement": "Staple food prices remained relatively stable from March 2019 to February 2020, with minor fluctuations.", "channels": ["wfp_staple_price"], "event_months": ["2019-03", "2020-02"], "source_url": null, "uncertainty": "Stable prices suggest stable market conditions, but do not directly indicate food security status."}], "rationale": "The food consumption score (FCS) and reduced coping strategy index (rCSI) both showed fluctuations over the observed period, with the FCS ending slightly lower and the rCSI slightly higher in February 2020 compared to March 2019. This suggests some instability in food consumption and coping strategies, but the normalization method is unknown, making it difficult to draw definitive conclusions about food security. Conflict events and fatalities decreased significantly, which could have a positive impact on food security by reducing disruptions. Rainfall was variable, with a peak in July 2019, which could affect agricultural productivity, but this is not directly measured. Staple food prices remained stable, suggesting stable market conditions. Overall, the observed data do not indicate a significant change in food security status, aligning with the stable IPC phase 2.", "uncertainty": "The normalization of FCS and rCSI indices is unknown, making it difficult to interpret their implications for food security. The indirect effects of conflict reduction and rainfall variability on food security are plausible but not directly measured.", "confidence": 0.8}}
|
| 2 |
+
{"sample_id": "0055737fcdedc492437f7aea", "dataset_version": "38c6ba2c21879552c4e03a7222462234a9a27c20a173d3adac4ffba77bc169a7", "purpose": "training_supervision_only", "request_sha256": "0856043bfdb6546ba328b86400015b34a068d7260b278919f68ee34f694e3267", "model": "gpt-4o-2024-08-06", "request_id": "chatcmpl-ENOWs9Tem95A0Umqmsts4TLqofvkd", "usage": {"completion_tokens": 948, "prompt_tokens": 2574, "total_tokens": 3522, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": null, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The fcs_rt mean index showed a general decline from November 2018 to March 2019, followed by fluctuations without a clear trend from April 2019 to October 2019.", "channels": ["fcs_rt mean"], "event_months": ["2018-11", "2018-12", "2019-01", "2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "The normalization method is unknown, so the severity of food consumption cannot be directly inferred."}, {"provenance": "observed", "statement": "The rcsi_rt mean index generally increased from November 2018 to October 2019, indicating more frequent use of coping strategies.", "channels": ["rcsi_rt mean"], "event_months": ["2018-11", "2018-12", "2019-01", "2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "The normalization method is unknown, so the severity of coping behavior cannot be directly inferred."}, {"provenance": "observed", "statement": "Portwatch import data showed fluctuations in dry bulk and cargo imports from January 2019 to October 2019, with no consistent upward or downward trend.", "channels": ["portwatch_import_dry_bulk", "portwatch_import_cargo"], "event_months": ["2019-01", "2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "Fluctuations in import data may affect food availability, but the direct impact on food security is unclear."}, {"provenance": "observed", "statement": "Conflict events and fatalities generally decreased from November 2018 to October 2019.", "channels": ["acled_events", "acled_fatalities"], "event_months": ["2018-11", "2018-12", "2019-01", "2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "Reduced conflict might improve food security, but other factors could offset this potential improvement."}, {"provenance": "observed", "statement": "Rainfall was variable with a notable increase in July and August 2019, followed by a decrease in September 2019.", "channels": ["chirps_rainfall"], "event_months": ["2018-11", "2018-12", "2019-01", "2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09"], "source_url": null, "uncertainty": "Variable rainfall can impact agricultural productivity, but the specific effects on food security are uncertain."}], "rationale": "The observed future phase remained at IPC phase 3, indicating no change in the food security situation. The fcs_rt mean index showed a general decline initially, suggesting potential stress in food consumption, but later fluctuations without a clear trend make it difficult to ascertain a direct impact on food security. The rcsi_rt mean index generally increased, which could imply increased stress and reliance on coping mechanisms. However, the normalization method for these indices is unknown, so direct severity cannot be inferred. Import data showed fluctuations, which might affect food availability but did not show a consistent trend. Conflict events and fatalities decreased, which could potentially improve food security conditions. Rainfall variability might affect agricultural productivity, but its direct impact on food security is uncertain. Overall, the mixed signals from various indicators suggest a complex food security environment without a clear deterioration or improvement, aligning with the observed phase stability.", "uncertainty": "The normalization methods for the FCS and rCSI indices are unknown, making it difficult to directly interpret the severity of food consumption and coping behavior. Additionally, the impact of fluctuating imports and variable rainfall on food security is uncertain, as these factors interact with other socio-economic and environmental conditions.", "confidence": 0.8}}
|
| 3 |
+
{"sample_id": "008741623f7b301f71537290", "dataset_version": "38c6ba2c21879552c4e03a7222462234a9a27c20a173d3adac4ffba77bc169a7", "purpose": "training_supervision_only", "request_sha256": "5c577364a5fc9f540c68b58afddbdc1fc6aedfdac20d26026f61a078de5dc060", "model": "gpt-4o-2024-08-06", "request_id": "chatcmpl-ENOWzOtH4BvybmLzLDpMt4pvoWdtG", "usage": {"completion_tokens": 718, "prompt_tokens": 2566, "total_tokens": 3284, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": null, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The fcs_rt mean index showed fluctuations with a general increase from June to October 2019.", "channels": ["fcs_rt mean"], "event_months": ["2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "The normalization method of the index is unknown, so the implications for food consumption are unclear."}, {"provenance": "observed", "statement": "The rcsi_rt mean index showed fluctuations with a general increase from June to October 2019.", "channels": ["rcsi_rt mean"], "event_months": ["2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "The normalization method of the index is unknown, so the implications for coping strategies are unclear."}, {"provenance": "observed", "statement": "There was a decrease in conflict events and fatalities from May to October 2019.", "channels": ["acled_events", "acled_fatalities"], "event_months": ["2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "The impact of reduced conflict on food security is plausible but not directly measurable from these indices."}, {"provenance": "observed", "statement": "Rainfall was relatively high in July and August 2019, followed by a decrease in September.", "channels": ["chirps_rainfall"], "event_months": ["2019-07", "2019-08", "2019-09"], "source_url": null, "uncertainty": "The effect of rainfall on agricultural productivity and food security is inferred but not directly measured."}, {"provenance": "observed", "statement": "Staple food prices remained relatively stable from February to October 2019.", "channels": ["wfp_staple_price"], "event_months": ["2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "Stable prices suggest stable market conditions, but the impact on food access is not directly measured."}], "rationale": "The observed data show that the fcs_rt mean index fluctuated but generally increased from June to October 2019, while the rcsi_rt mean index also showed an increasing trend over the same period. This suggests potential stability or improvement in food consumption and coping strategies, although the exact implications are unclear due to unknown index normalization. Conflict events and fatalities decreased from May to October 2019, which could have contributed to improved security and stability, potentially benefiting food security. Rainfall was high in July and August, which might have supported agricultural activities, but its direct impact on food security is not measured. Staple food prices remained stable, indicating steady market conditions. Despite these factors, the IPC phase remained stable at 3, suggesting that while some indicators showed positive trends, they were not sufficient to improve the overall food security classification. The uncertainty lies in the indirect nature of these indicators and their complex interactions.", "uncertainty": "The normalization of indices and indirect nature of indicators introduce uncertainty in interpreting their impact on food security.", "confidence": 0.8}}
|
annotations/annotations-pilot-v2.manifest.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"requested": 3,
|
| 3 |
+
"valid": 3,
|
| 4 |
+
"failures": [],
|
| 5 |
+
"model": "gpt-4o-2024-08-06",
|
| 6 |
+
"conservative_cost_bound_usd": 0.11265,
|
| 7 |
+
"estimated_usage_usd": 0.0428425,
|
| 8 |
+
"pricing_source": "https://developers.openai.com/api/docs/models/gpt-4o",
|
| 9 |
+
"cached_usage_may_include_previous_runs": true
|
| 10 |
+
}
|
annotations/annotations-pilot.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"sample_id": "001e6338a8099dd54cf35585", "dataset_version": "38c6ba2c21879552c4e03a7222462234a9a27c20a173d3adac4ffba77bc169a7", "purpose": "training_supervision_only", "request_sha256": "c31d7c1396cc9fee1b61d630a4948d8e7bd98fa09b209a1ae913ace764cb1ff9", "model": "gpt-4o-2024-08-06", "request_id": "chatcmpl-ENOUwQM6ReFqv1BxpYTOxUQAN2lK4", "usage": {"completion_tokens": 441, "prompt_tokens": 2516, "total_tokens": 2957, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": null, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 2, "precursor_patterns": [{"provenance": "observed", "statement": "The FCS index remained relatively stable with minor fluctuations, showing no clear trend of improvement or deterioration.", "channels": ["fcs_rt mean"], "event_months": ["2019-03", "2020-02"], "source_url": null, "uncertainty": "The FCS values fluctuated but did not show a clear upward or downward trend."}, {"provenance": "observed", "statement": "The rCSI index showed fluctuations but no consistent trend, with a peak in January 2020.", "channels": ["rcsi_rt mean"], "event_months": ["2019-03", "2020-02"], "source_url": null, "uncertainty": "The rCSI values fluctuated, indicating varying levels of coping strategies without a clear trend."}, {"provenance": "observed", "statement": "There was a general decline in conflict events and fatalities, with a significant drop from mid-2019 to early 2020.", "channels": ["acled_events", "acled_fatalities"], "event_months": ["2019-03", "2020-02"], "source_url": null, "uncertainty": "The decline in conflict events and fatalities could contribute to food security stability."}, {"provenance": "observed", "statement": "Rainfall was variable with no clear pattern, and there was missing data for February 2020.", "channels": ["chirps_rainfall"], "event_months": ["2019-03", "2020-02"], "source_url": null, "uncertainty": "Variable rainfall patterns can affect agricultural productivity and food security."}], "rationale": "The food security phase remained stable at 2, indicating no significant improvement or deterioration. The FCS and rCSI indices showed fluctuations but no consistent trends, suggesting stability in food consumption and coping strategies. The decline in conflict events and fatalities may have contributed to this stability. However, variable rainfall and missing data for February 2020 introduce uncertainty regarding agricultural conditions.", "uncertainty": "The stability in food security is uncertain due to variable rainfall patterns and missing data, which could affect agricultural productivity.", "confidence": 0.85}}
|
| 2 |
+
{"sample_id": "0055737fcdedc492437f7aea", "dataset_version": "38c6ba2c21879552c4e03a7222462234a9a27c20a173d3adac4ffba77bc169a7", "purpose": "training_supervision_only", "request_sha256": "f4515a6a30ebd59c89954382394e7bd100e17378451de068f543bed9b8718f76", "model": "gpt-4o-2024-08-06", "request_id": "chatcmpl-ENOV14rUb3na281KfZVp4dricqm8h", "usage": {"completion_tokens": 639, "prompt_tokens": 2497, "total_tokens": 3136, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": null, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The FCS index showed a general decline from November 2018 to May 2019, with some fluctuations thereafter.", "channels": ["fcs_rt mean"], "event_months": ["2018-11", "2018-12", "2019-01", "2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "The exact impact of FCS changes on food security phase is uncertain due to unknown normalization."}, {"provenance": "observed", "statement": "The rCSI index increased from November 2018 to October 2019, indicating increased coping strategies.", "channels": ["rcsi_rt mean"], "event_months": ["2018-11", "2018-12", "2019-01", "2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "The relationship between rCSI and food security phase is not straightforward due to normalization."}, {"provenance": "observed", "statement": "There was a significant increase in dry bulk and cargo imports from February to April 2019, followed by fluctuations.", "channels": ["portwatch_import_dry_bulk", "portwatch_import_cargo"], "event_months": ["2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "The impact of import changes on local food security is uncertain without detailed distribution data."}, {"provenance": "observed", "statement": "Conflict events and fatalities decreased from November 2018 to October 2019.", "channels": ["acled_events", "acled_fatalities"], "event_months": ["2018-11", "2018-12", "2019-01", "2019-02", "2019-03", "2019-04", "2019-05", "2019-06", "2019-07", "2019-08", "2019-09", "2019-10"], "source_url": null, "uncertainty": "The direct impact of conflict reduction on food security is complex and mediated by other factors."}], "rationale": "The FCS and rCSI indices suggest worsening food consumption and increased coping strategies, respectively. However, the increase in imports and decrease in conflict events and fatalities may have mitigated potential deterioration in food security. The observed phase stability could be due to these mitigating factors, despite negative trends in food consumption indicators.", "uncertainty": "The normalization of FCS and rCSI indices is unknown, making their direct interpretation challenging. The impact of imports and conflict on food security is also complex and context-dependent.", "confidence": 0.8}}
|
| 3 |
+
{"sample_id": "008741623f7b301f71537290", "dataset_version": "38c6ba2c21879552c4e03a7222462234a9a27c20a173d3adac4ffba77bc169a7", "purpose": "training_supervision_only", "request_sha256": "d5d30d8aa532d43d71cc395082a89aa3e259eb0331a6a12b03a028fe0d7a3d5b", "model": "gpt-4o-2024-08-06", "request_id": "chatcmpl-ENOV7tueUGnRQZHLLBrYNltOwCUc5", "usage": {"completion_tokens": 457, "prompt_tokens": 2489, "total_tokens": 2946, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": null, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The FCS index showed a decline from November 2018 to June 2019, followed by an increase from July 2019 to October 2019.", "channels": ["fcs_rt mean"], "event_months": ["2018-11", "2019-10"], "source_url": null, "uncertainty": "The normalization method is unknown, so the severity of changes is uncertain."}, {"provenance": "observed", "statement": "The rCSI index decreased from November 2018 to February 2019, then fluctuated with a slight increase towards October 2019.", "channels": ["rcsi_rt mean"], "event_months": ["2018-11", "2019-10"], "source_url": null, "uncertainty": "The normalization method is unknown, so the severity of changes is uncertain."}, {"provenance": "observed", "statement": "There was a high level of conflict events and fatalities from November 2018 to April 2019, with a decrease in both metrics towards October 2019.", "channels": ["acled_events", "acled_fatalities"], "event_months": ["2018-11", "2019-10"], "source_url": null, "uncertainty": "The impact of conflict on food security is complex and not directly quantifiable from these metrics alone."}, {"provenance": "observed", "statement": "Port imports of dry bulk and cargo were variable, with peaks in April 2019 and fluctuations thereafter.", "channels": ["portwatch_import_dry_bulk", "portwatch_import_cargo"], "event_months": ["2019-01", "2019-10"], "source_url": null, "uncertainty": "The relationship between import levels and local food security is not directly clear."}], "rationale": "The FCS and rCSI indices showed fluctuations with some recovery towards the end of the period. Conflict events and fatalities were high initially but decreased over time. Imports showed variability but no clear trend indicating a crisis. These mixed signals suggest stability in the food security phase.", "uncertainty": "The normalization of indices and complex interactions between conflict, imports, and food security contribute to uncertainty.", "confidence": 0.8}}
|
annotations/annotations-smoke-train.jsonl
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"sample_id": "0077e979b13f48c0cb8c6c50", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "0abe933634c534cde064bcc07cebdd5f7f91e8b1b4074d291dda6d0acb0d82e5", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhhruVstcBHM0v3NI71oX5LI3OF", "usage": {"completion_tokens": 1438, "prompt_tokens": 4311, "total_tokens": 5749, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 114, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 4308, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The most recent IPC assessment available at the July 2020 cutoff was Phase 2, with assessment event month June 2020 and an age of 1 month at cutoff.", "channels": ["last_available_phase"], "event_months": ["2020-06"], "source_url": null, "uncertainty": "This is the latest supplied assessment date, not evidence that a new assessment occurred on the cutoff date."}, {"provenance": "observed", "statement": "Country-level conflict events rose from 787 in January to 945 in March and 962 in April; fatalities peaked at 2,115 in April. The rolling three-month event total was 2,711 in April and 2,732 in May.", "channels": ["acled_events", "acled_fatalities", "acled_events_sum_3m"], "event_months": ["2020-03", "2020-04", "2020-05"], "source_url": null, "uncertainty": "Conflict measures are national aggregates and do not establish exposure or impacts in Hajr."}, {"provenance": "observed", "statement": "The national imported-rice price was USD 3.61/kg from January through April, then rose 3.05% to USD 3.72/kg in May and remained USD 3.72/kg in June.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2020-05", "2020-06"], "source_url": null, "uncertainty": "The price is a country-level series for one imported staple and may not represent prices or household purchasing conditions in Hajr."}, {"provenance": "observed", "statement": "National cargo imports declined from 770,018 metric tons in March to 675,759 in April and 629,638 in May; the supplied comparison with the preceding three-month mean was -6.65% in April and -12.50% in May, before imports rose to 881,028 tons in June.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct"], "event_months": ["2020-03", "2020-04", "2020-05", "2020-06"], "source_url": null, "uncertainty": "Cargo imports are not food-specific, are national rather than district-level, and the June rebound limits evidence of a sustained import decline."}, {"provenance": "observed", "statement": "The admin2 FCS normalized index fell by 0.0350 from April to May and then rose by 0.0743 in June; the rCSI normalized index declined by 0.0158 from April to May and changed by +0.0026 in June.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "rcsi_rt mean", "rcsi_rt mean_change_1m"], "event_months": ["2020-04", "2020-05", "2020-06"], "source_url": null, "uncertainty": "HFID normalization direction and method are unknown, so these numeric movements cannot be interpreted as changes in food consumption or coping severity."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "The April conflict peak (962 events and 2,115 fatalities) preceded the May imported-rice price increase of 3.05%, while national cargo imports were below their previous-three-month comparison in April (-6.65%) and May (-12.50%).", "mechanism": "Heightened conflict can plausibly disrupt transport, market operations, and household income, while reduced aggregate import throughput may constrain wider market availability; together these pressures could contribute to higher staple costs and weaker purchasing access. This is a plausible compounding mechanism, not a proven causal chain, and cargo is not food-specific.", "channels": ["acled_events", "acled_fatalities", "portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "wfp_staple_price", "wfp_staple_price_change_1m_pct"], "event_months": ["2020-04", "2020-05"], "source_url": null, "uncertainty": "The signals are country-level and temporally associated only; neither local market transmission nor causality in Hajr is observed."}, {"provenance": "inference", "statement": "The June 2020 Phase 2 assessment was followed by the observed October 2020 Phase 3 outcome, after March-May elevated national conflict activity and a May rise in the imported-rice price.", "mechanism": "Persistent conflict-related disruption and higher staple costs could plausibly erode household access over subsequent months, helping explain a transition from Phase 2 to Phase 3. The assessed phase change is the evidence of food-security deterioration; the normalized FCS/rCSI values are not used to infer its direction.", "channels": ["last_available_phase", "acled_events_sum_3m", "wfp_staple_price", "wfp_staple_price_change_3m_pct"], "event_months": ["2020-03", "2020-04", "2020-05", "2020-06", "2020-10"], "source_url": null, "uncertainty": "The October result may reflect unobserved local shocks, livelihoods, assistance, and assessment changes; this retrospective hypothesis does not demonstrate attribution."}], "rationale": "Observed ground truth is Phase 3 in October 2020, a one-phase deterioration from the latest available June Phase 2 assessment. Before that outcome, national conflict intensity was elevated in March-May, imported-rice prices increased in May, and cargo imports fell in April-May before recovering in June. These are plausible warning signals but are mostly national aggregates. Admin2 FCS and rCSI show numeric fluctuations only; because their normalization is undocumented, they are not interpreted as food-consumption or coping changes. The IPC phase change, rather than those indexes, establishes deterioration.", "uncertainty": "Evidence is short-windowed, mostly national in scope, and cannot identify causal pathways to Hajr. Rainfall has no supplied anomaly baseline and June is missing, so it is not interpreted. The June assessment was recent at cutoff but may not capture conditions that developed before October.", "confidence": 0.72}}
|
| 2 |
+
{"sample_id": "000fa9a9ece5eae8cf8e209d", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "458836a3800af43cb1a31210215fec4f6b2964fa87a728c79297918b3e3bfc2a", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhhpK6d2NvhhZJnb3nTZU4Q267J", "usage": {"completion_tokens": 1091, "prompt_tokens": 4189, "total_tokens": 5280, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 70, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 2854, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 1, "precursor_patterns": [{"provenance": "observed", "statement": "The latest available IPC assessment was Phase 1 in 2022-02; the rolling last-known Phase 1 observations were tied to 2021-06 and 2021-10 assessment events, rather than new assessments in every series month.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2021-06", "2021-10", "2022-02"], "source_url": null, "uncertainty": "The rolling assessment-age values describe information available at each historical month. The latest assessment was dated 2022-02 and was one month old at the 2022-03 cutoff."}, {"provenance": "observed", "statement": "Country-level maize price rose from USD 0.20/kg in 2021-12 to USD 0.25/kg in 2022-01 and USD 0.26/kg in 2022-02; reported one-month changes were +25.0% and +4.0%, and the three-month change reached +36.8% in 2022-02.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2021-12", "2022-01", "2022-02"], "source_url": null, "uncertainty": "These are country-level staple-price observations and do not demonstrate prices, access, or purchasing power in SC Chitekwele."}, {"provenance": "observed", "statement": "Admin2 HFID normalized FCS values varied between 0.121 and 0.174 during 2021-09 to 2022-02, ending with a +0.033 one-month numeric change in 2022-02; normalized rCSI varied between 0.291 and 0.329 and had a -0.037 one-month numeric change in 2022-02.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "rcsi_rt mean", "rcsi_rt mean_change_1m"], "event_months": ["2021-09", "2022-02"], "source_url": null, "uncertainty": "HFID normalization and index direction are undocumented, so these numeric movements cannot be interpreted as changes in food consumption or coping behavior."}, {"provenance": "observed", "statement": "Country rainfall increased from 4.0 mm in 2021-09 to 347.7 mm in 2022-01 after intermediate monthly totals of 15.3, 47.8, and 54.8 mm; 2022-02 rainfall was missing.", "channels": ["chirps_rainfall"], "event_months": ["2021-09", "2022-01"], "source_url": null, "uncertainty": "Rainfall is country-level, lacks a climatological benchmark, and therefore cannot establish whether conditions in the district were favorable or adverse."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "The country-level maize-price rise in 2022-01 to 2022-02 coincided with the district's latest Phase 1 assessment event in 2022-02.", "mechanism": "If the national price increase was transmitted to SC Chitekwele markets, higher staple costs could have strained household purchasing power; the Phase 1 assessment indicates that no phase deterioration was recorded at that assessment. This timing is a plausible market-access consideration, not evidence that prices caused or did not cause local food-security outcomes.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "last_available_phase"], "event_months": ["2022-01", "2022-02"], "source_url": null, "uncertainty": "The price series is national and the IPC result is an area classification; neither identifies household exposure, market pass-through, income, or causality."}], "rationale": "The supplied future outcome is IPC Phase 1 in 2022-06, unchanged from the latest Phase 1 assessment dated 2022-02. Before the cutoff, measured signals include a late increase in the national maize-price series, fluctuating but direction-ambiguous normalized FCS/rCSI values, and rising country rainfall totals through 2022-01. These patterns provide limited contextual precursors but do not override the directly supplied IPC assessment trajectory or establish a causal pathway.", "uncertainty": "Evidence is limited by short series, missing February rainfall and all trade/shipping data, national rather than district coverage for price, conflict, and rainfall channels, and unknown HFID index normalization. IPC assessment timing is explicit: the current assessment event is 2022-02, not the age values attached to earlier rolling observations.", "confidence": 0.8}}
|
| 3 |
+
{"sample_id": "002484ea2c5aaa22f62d46e8", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "044ebd6626c50ade6599eeb9dc88c365fed15ee03f525100161392823fc47aa0", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhhAq5XSOOuzKLB8NmuMQzMEyzU", "usage": {"completion_tokens": 1490, "prompt_tokens": 4209, "total_tokens": 5699, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 320, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 2874, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 1, "precursor_patterns": [{"provenance": "observed", "statement": "The national white-maize price rose from USD 0.41/kg in March to 0.47/kg in April, then was USD 0.45/kg in May and USD 0.46/kg in June; the reported three-month changes were +19.0%, +15.4%, and +12.2% in April\u2013June.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This is a country-level monitored staple-price series and may not represent prices, market access, or purchasing power in Oubritenga."}, {"provenance": "observed", "statement": "Country-level ACLED events peaked at 191 in February and were 135 in June; the trailing three-month event total declined from 511 in March to 402 in June. Fatalities fluctuated, reaching 439 in May before falling to 363 in June.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-02", "2022-03", "2022-05", "2022-06"], "source_url": null, "uncertainty": "National event and fatality counts do not measure exposure to conflict within Oubritenga and may reflect reporting variation."}, {"provenance": "observed", "statement": "Reported country-level rainfall increased sequentially from 0.36 mm in January to 56.27 mm in May; June rainfall is missing.", "channels": ["chirps_rainfall"], "event_months": ["2022-01", "2022-02", "2022-03", "2022-04", "2022-05"], "source_url": null, "uncertainty": "The series is national rather than district-specific, and the short seasonal sequence lacks an anomaly baseline."}, {"provenance": "observed", "statement": "The admin2 normalized FCS series declined from 0.679 in February to 0.626 in May before increasing to 0.663 in June. The normalized rCSI series increased from 0.130 in April to 0.195 in May and 0.197 in June.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "rcsi_rt mean", "rcsi_rt mean_change_1m"], "event_months": ["2022-02", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID normalization and index direction are undocumented; these numeric movements cannot be interpreted as changes in food consumption or coping behavior."}, {"provenance": "observed", "statement": "Rolling IPC entries through June retained phase 1, with the underlying assessment event updating from October 2021 to February 2022. Separately, the latest available assessment at the July cutoff was phase 2 with event month June 2022 and age one month at cutoff.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2021-10", "2022-02", "2022-06"], "source_url": null, "uncertainty": "The rolling ages apply to their respective as-of months, whereas the latest assessment has a distinct June event month and one-month age at cutoff."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "April\u2013June national maize prices remained 12.2%\u201319.0% above their respective values three months earlier while national conflict events and fatalities remained substantial (120\u2013147 events and 363\u2013439 fatalities per month in April\u2013June).", "mechanism": "Elevated staple prices could constrain purchasing power, while conflict can plausibly disrupt livelihoods, movement, or local market access; together these pressures could compound household difficulty obtaining food. This is a plausible multi-domain hypothesis rather than evidence that either national signal caused the later IPC phase change in Oubritenga.", "channels": ["wfp_staple_price", "wfp_staple_price_change_3m_pct", "acled_events", "acled_fatalities"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "Both price and conflict measures are national aggregates, and neither establishes local exposure, transmission mechanisms, or causation."}], "rationale": "The supplied outcome is IPC phase 1 in October 2022, with phase_change -1 from the latest known phase 2 assessment dated June 2022. Before that outcome, measured signals showed sustained year-on-quarter maize-price increases, declining but still substantial national conflict activity, rising January\u2013May national rainfall, and mixed numeric movements in the normalized admin2 FCS and rCSI series. The index movements are reported only numerically because their normalization direction is unknown. The observed phase transition, rather than these correlations, is the evidence for food-security change.", "uncertainty": "Evidence is limited to six months, rainfall is missing for June, and trade channels are entirely unavailable. Conflict, rainfall, and price series are national and cannot be assumed to characterize Oubritenga. The newest IPC assessment is June 2022, one month old at the July cutoff; it is distinct from the rolling assessment-age entries. Associations and plausible mechanisms do not demonstrate causation.", "confidence": 0.66}}
|
| 4 |
+
{"sample_id": "00147bb957ae7aee24311f68", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "21ca5d3a2130544b572accd32712142f3d60e09c519668cd4bf340639829bf7c", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhsfzgx0iSm14vXcPsS82rTYIav", "usage": {"completion_tokens": 1434, "prompt_tokens": 4241, "total_tokens": 5675, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 2906, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 2, "precursor_patterns": [{"provenance": "observed", "statement": "The IPC phase recorded for San Antonio Huista remained Phase 2 throughout the rolling January-June window. The underlying assessment event changed from October 2021 to February 2022 in March; the latest assessment available at the July cutoff was also Phase 2, with event month June 2022 and an age of one month at cutoff.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2021-10", "2022-02", "2022-06"], "source_url": null, "uncertainty": "Rolling IPC ages apply to their respective as-of months, while the June assessment and its one-month age at cutoff are separately supplied. These observations establish assessment continuity, not conditions between assessments."}, {"provenance": "observed", "statement": "The admin2 normalized FCS series declined from 0.1291 in January to 0.0703 in June, with negative three-month changes in April (-0.0192), May (-0.0394), and June (-0.0481).", "channels": ["fcs_rt mean", "fcs_rt mean_change_3m"], "event_months": ["2022-01", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID normalization and direction of severity are undocumented; this is only a numeric index decline and cannot by itself be characterized as a food-consumption change."}, {"provenance": "observed", "statement": "The admin2 normalized rCSI series fell from 0.2819 in January to 0.1955 in June. Three-month differences were negative in April (-0.0542), May (-0.1219), and June (-0.0716), although the series rose by 0.0186 from May to June.", "channels": ["rcsi_rt mean", "rcsi_rt mean_change_1m", "rcsi_rt mean_change_3m"], "event_months": ["2022-01", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "The unknown HFID normalization precludes interpreting the numeric movement as a change in coping behavior."}, {"provenance": "observed", "statement": "Country-level cargo imports rose from 956,198 metric tons in April to 1,123,380 in June; imports were 2.7%, 13.9%, and 12.0% above their respective previous-three-month means in April, May, and June. June dry-bulk imports were 913,751 metric tons.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "portwatch_import_dry_bulk"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "Cargo and dry-bulk imports are not food-specific and are national rather than district measures; these data do not demonstrate local availability or access."}, {"provenance": "observed", "statement": "Country-level conflict-event counts were lower late in the window: the three-month event sum declined from 227 in March to 203 in April and May, then 188 in June. Monthly fatalities varied from 46 in April to 66 in June.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "These are country totals and cannot establish exposure, displacement, or market disruption in San Antonio Huista."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "From April through June, country cargo imports were above their prior-three-month averages while the admin2 FCS and rCSI normalized series showed negative three-month differences; the latest June IPC assessment nevertheless remained Phase 2.", "mechanism": "If part of the increased national cargo flow supported market supply, it may have buffered some supply-side pressures, but national non-food-specific imports need not reach this district or overcome household access constraints. The Phase 2 assessment is consistent with no documented phase change, but the combined signals do not prove buffering or any causal relationship.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "fcs_rt mean_change_3m", "rcsi_rt mean_change_3m", "last_available_phase"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This combines national trade data, district normalized indices of unknown direction, and an IPC assessment; neither market transmission nor food content of imports is observed."}, {"provenance": "inference", "statement": "The country three-month conflict-event total decreased from 227 in March to 188 in June while May rainfall reached 200.0 mm and the June assessment recorded IPC Phase 2 in San Antonio Huista.", "mechanism": "Reduced aggregate conflict activity and seasonal rainfall could plausibly lessen disruption to livelihoods, transport, or agricultural activity, helping support stability in food-security conditions. This is a possible compounding context rather than evidence that either national signal caused the district's Phase 2 outcome.", "channels": ["acled_events_sum_3m", "chirps_rainfall", "last_available_phase"], "event_months": ["2022-03", "2022-05", "2022-06"], "source_url": null, "uncertainty": "Conflict and rainfall are country-level aggregates, June rainfall is missing, and no local production, displacement, market, or household-access observations are supplied."}], "rationale": "Observed outcome: IPC Phase 2 in October 2022, unchanged from the June 2022 assessment (Phase 2). Before the outcome, the rolling IPC record was consistently Phase 2 and the cutoff assessment was recent. Measured numeric patterns include declines in both normalized admin2 FCS and rCSI series over much of January-June, increased national cargo imports relative to earlier three-month averages late in the period, and a declining national three-month conflict-event count. Because the index normalizations have unknown severity direction, their movements are reported only numerically and do not identify food-consumption or coping changes. The assessment evidence\u2014not those indices\u2014supports the documented absence of an IPC phase change. Plausible cross-domain mechanisms remain unproven, particularly given national-versus-district mismatch and missing staple-price data.", "uncertainty": "No staple-price observations are available. Trade, conflict, and rainfall signals are national aggregates; cargo measures are not food-specific. Rainfall is missing for June. The October target assessment is not supplied beyond its ground-truth phase, so intervening causal processes cannot be established.", "confidence": 0.78}}
|
| 5 |
+
{"sample_id": "001f457a9d18ffae84ae0525", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "33fdb3fe5de2854afad50ed4f9a58a23ed606a5a2b3e5101a73f7bacdaf41b9d", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhvALSg1UCEPXSGFFcL4TpJsJPw", "usage": {"completion_tokens": 1413, "prompt_tokens": 4340, "total_tokens": 5753, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 3005, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 2, "precursor_patterns": [{"provenance": "observed", "statement": "The latest IPC assessment was phase 2 in 2022-06, one month old at the 2022-07 cutoff; the prior rolling phase-1 values referred to assessments from 2021-10 and 2022-02 rather than new monthly assessments.", "channels": ["last_available_phase", "ipc_last_known_phase", "ipc_assessment_age"], "event_months": ["2021-10", "2022-02", "2022-06"], "source_url": null, "uncertainty": "IPC month stamps are assessment event months. The rolling phase series is stale-status information, while the separate last_available_phase identifies the latest assessment."}, {"provenance": "observed", "statement": "The country-level millet price rose from USD 1.385/2.6 kg in 2022-01 to 1.525 in 2022-04 (a 10.1% three-month increase), then was USD 1.475 in 2022-06.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2022-01", "2022-04", "2022-06"], "source_url": null, "uncertainty": "This is a country-level quoted staple price and may not represent prices or household purchasing conditions in Okehi."}, {"provenance": "observed", "statement": "Country-level conflict-event counts declined overall from 353 in 2022-01 to 299 in 2022-06; the trailing three-month total fell from 1,019 in 2022-03 to 892 in 2022-06. Fatalities varied, peaking at 1,571 in 2022-03 before reaching 809 in 2022-06.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-01", "2022-03", "2022-06"], "source_url": null, "uncertainty": "Country aggregates do not establish exposure, displacement, or market disruption in Okehi."}, {"provenance": "observed", "statement": "Okehi's normalized FCS series fell by 0.0461 in 2022-02, then rose by 0.0294 in 2022-05 and 0.0013 in 2022-06; the normalized rCSI series fell by 0.0713 in 2022-03, rose in April-May, and fell by 0.0258 in June.", "channels": ["fcs_rt mean", "fcs_rt mean_change_1m", "rcsi_rt mean", "rcsi_rt mean_change_1m"], "event_months": ["2022-02", "2022-03", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID normalization and index direction are undocumented, so these numeric movements cannot be interpreted as changes in food consumption or coping behavior."}, {"provenance": "observed", "statement": "Country-level rainfall increased from 3.5 mm in 2022-02 to 25.3 mm in March, 81.2 mm in April, and 109.1 mm in May.", "channels": ["chirps_rainfall"], "event_months": ["2022-02", "2022-03", "2022-04", "2022-05"], "source_url": null, "uncertainty": "No June rainfall value is available, and country-level rainfall does not measure local agricultural outcomes in Okehi."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "The 10.1% country-level three-month millet-price increase to 2022-04 coincided with the onset of markedly higher country-level rainfall in March-May, while the later Okehi phase-2 assessment was dated 2022-06.", "mechanism": "Higher staple prices can plausibly constrain market purchasing power during a seasonal transition in which household reliance on markets and agricultural conditions may be changing. This combination could have contributed to food-security stress reflected in the June phase-2 assessment, but the supplied data do not demonstrate local household exposure, seasonal dependence, or causation.", "channels": ["wfp_staple_price", "wfp_staple_price_change_3m_pct", "chirps_rainfall", "last_available_phase"], "event_months": ["2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "Price and rainfall are country-level signals, rainfall alone does not identify production effects, and the phase assessment cannot be causally attributed to these concurrent patterns."}, {"provenance": "inference", "statement": "Country conflict events remained high in absolute number (892 over April-June) while import cargo fell 8.35% relative to the prior three-month mean in 2022-06; Okehi's most recent assessment was phase 2 in June.", "mechanism": "Persistent conflict can plausibly impede domestic trade and market access, and reduced national cargo imports could plausibly tighten broader logistics or availability. Together they might compound local market risks, but cargo is not food-specific and neither national measure establishes a supply shock or conflict impact in Okehi.", "channels": ["acled_events_sum_3m", "portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "last_available_phase"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This is a plausible, unproven cross-domain mechanism. National cargo and conflict aggregates cannot be treated as district-level market or household conditions."}], "rationale": "Observed outcome: IPC phase 2 in 2022-10, unchanged from the supplied latest June 2022 phase-2 assessment. Retrospectively, the main measurable lead signals were an early-year national millet-price rise, seasonal rainfall increase through May, and persistently high but declining national conflict-event totals. Okehi's normalized FCS and rCSI values changed numerically but have unknown normalization/direction and therefore do not support an interpretation of food-consumption or coping change. The available evidence is consistent with continued phase 2, but associations across national and district series do not establish causal drivers.", "uncertainty": "The target assessment basis is not supplied. Most prospective channels are national rather than Okehi-specific; import cargo and dry bulk are not food-specific; and the HFID index orientations are unknown. The June IPC assessment is the relevant latest assessment, not an extrapolation from the earlier rolling IPC values.", "confidence": 0.68}}
|
| 6 |
+
{"sample_id": "0029fa7cda0aa0fd84bee13c", "dataset_version": "8142c89de89cb862117d6a814be515b9f00312bb960e33cf7e994a99ed124bd3", "purpose": "training_supervision_only", "request_sha256": "c002fdf01b001791520bf080a6d4bec7c1232b23e5b0f18cd7561309384d602b", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENPhw4pMRohGUTOENzlS8kSbe1WaS", "usage": {"completion_tokens": 1445, "prompt_tokens": 4321, "total_tokens": 5766, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 67, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 2986, "cached_tokens": 1332, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The Beni IPC phase was 3 throughout January\u2013June 2022. The rolling values in January\u2013February refer to an October 2021 assessment, while those in March\u2013June refer to a February 2022 assessment; the latest available assessment at the July cutoff was also phase 3, with event month June 2022 and age one month at cutoff.", "channels": ["ipc_last_known_phase", "ipc_assessment_age", "last_available_phase"], "event_months": ["2021-10", "2022-02", "2022-06"], "source_url": null, "uncertainty": "The rolling IPC series is a record of the latest assessment available in each month, not evidence of a new assessment in every displayed month."}, {"provenance": "observed", "statement": "Cassava-flour price increased from USD 0.40/kg in January to USD 0.49/kg in June, after a low of USD 0.35/kg in March. The three-month price changes were +20.0% in April, +16.3% in May, and +40.0% in June.", "channels": ["wfp_staple_price", "wfp_staple_price_change_1m_pct", "wfp_staple_price_change_3m_pct"], "event_months": ["2022-01", "2022-03", "2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This is a country-level price series for one staple product and may not represent prices or household market access in Beni."}, {"provenance": "observed", "statement": "Country-level ACLED event counts fell from 561 in January to 393 in March, then rose to 513 in June; the trailing three-month event total was 1,407 in June. Reported fatalities declined overall from 994 in January to 682 in June.", "channels": ["acled_events", "acled_events_sum_3m", "acled_fatalities"], "event_months": ["2022-01", "2022-03", "2022-06"], "source_url": null, "uncertainty": "Country totals do not measure exposure, displacement, or access conditions in Beni specifically."}, {"provenance": "observed", "statement": "National cargo imports were 63,312 metric tons in April, then 115,041 in May and 128,635 in June; the import-cargo comparison with the previous three months was -24.1% in April, +44.4% in May, and +29.6% in June. Cargo imports are not food-specific.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "National port cargo volumes cannot establish food availability, commodity composition, inland delivery, or market supply in Beni."}, {"provenance": "observed", "statement": "The normalized FCS series rose from 0.3815 in April to 0.4361 in June (+0.0448 over three months), while the normalized rCSI series fell from 0.4692 to 0.4116 (-0.0655 over three months).", "channels": ["fcs_rt mean", "fcs_rt mean_change_3m", "rcsi_rt mean", "rcsi_rt mean_change_3m"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "HFID normalization method and severity direction are undocumented; these numeric movements cannot be interpreted as changes in food consumption or coping behavior."}], "interaction_hypotheses": [{"provenance": "inference", "statement": "The April\u2013June rise in the country-level cassava-flour price coincided with sustained high national conflict-event totals (1,287\u20131,407 events in trailing three-month windows) while Beni's latest June assessment remained IPC phase 3.", "mechanism": "Higher staple costs could plausibly strain purchasing power, while conflict can plausibly disrupt livelihoods, movement, and market access. Together these pressures could help maintain acute food insecurity rather than permit a phase improvement.", "channels": ["wfp_staple_price", "wfp_staple_price_change_3m_pct", "acled_events_sum_3m", "last_available_phase"], "event_months": ["2022-04", "2022-05", "2022-06"], "source_url": null, "uncertainty": "This is a plausible multi-domain mechanism, not a demonstrated causal chain: price and conflict data are national, whereas the IPC outcome is for Beni."}, {"provenance": "inference", "statement": "National cargo imports rebounded in May\u2013June as the cassava-flour price remained elevated relative to January and Beni's latest assessment was phase 3 in June.", "mechanism": "Even if some cargo movements supported aggregate supply, non-food cargo composition and constraints between ports and Beni could prevent national import increases from translating into lower local staple prices or improved household access.", "channels": ["portwatch_import_cargo", "portwatch_import_cargo_vs_previous_3m_pct", "wfp_staple_price", "last_available_phase"], "event_months": ["2022-05", "2022-06"], "source_url": null, "uncertainty": "Cargo is not food-specific, and neither import data nor the price series identifies Beni supply conditions; no causal effect can be inferred."}], "rationale": "The observed future phase is 3, unchanged from the latest June 2022 assessment available at the cutoff. Retrospectively, the record shows persistence of phase 3 alongside rising country-level cassava-flour prices and continued high national conflict-event counts. National cargo imports recovered late in the window, but those data are not food-specific and do not measure availability in Beni. The FCS and rCSI series show opposite numeric directions in April\u2013June, but their undocumented normalization precludes interpretation as food-consumption or coping change. These correlated signals plausibly align with continued phase 3, without proving causation.", "uncertainty": "IPC timing is explicitly resolved, with the June assessment being the relevant latest assessment at cutoff. Most potential drivers are country-level rather than Beni-level, rainfall is missing for June, and normalized FCS/rCSI direction is unknown. Accordingly, mechanisms are tentative and the annotation does not attribute the unchanged phase to any single channel.", "confidence": 0.72}}
|
annotations/annotations-terra-pilot-final.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"sample_id": "000de9411dbce885b4c96ce5", "dataset_version": "08268e585e0e9028834bf68007a6fb710581b48da0a88ed556fb8502dbfec270", "purpose": "training_supervision_only", "request_sha256": "df668c8e6bf2d1c69d3fb908c2dcb8489734a090fff756932e3b1574b2a055a8", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENOjhQ8y1xUD5B8v3RA8K1znKhGln", "usage": {"completion_tokens": 896, "prompt_tokens": 1984, "total_tokens": 2880, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 1981, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 3, "precursor_patterns": [{"provenance": "observed", "statement": "The observed IPC phase was 3 in February 2022 and remained 3 in the June 2022 target month (phase change 0).", "channels": ["last_available_phase"], "event_months": ["2022-02", "2022-06"], "source_url": null, "uncertainty": "The supplied phase observations establish no phase change over the stated interval, but do not identify conditions within the intervening months."}, {"provenance": "observed", "statement": "Both normalized admin2 survey series declined overall before the cutoff: FCS from 0.548 in September to 0.401 in February, and rCSI from 0.730 to 0.562, with FCS partly rebounding after its November low of 0.369.", "channels": ["fcs_rt mean", "rcsi_rt mean"], "event_months": ["2021-09", "2021-10", "2021-11", "2021-12", "2022-01", "2022-02"], "source_url": null, "uncertainty": "Normalization method and direction are unknown; these are numeric movements only and cannot be interpreted as changes in food consumption or coping behavior."}, {"provenance": "observed", "statement": "Country-level conflict event counts rose from 252 in December to 353 in January and remained 332 in February; fatalities rose from 705 to 1,327 in January before falling to 891 in February.", "channels": ["acled_events", "acled_fatalities"], "event_months": ["2021-12", "2022-01", "2022-02"], "source_url": null, "uncertainty": "These are country-level aggregates and may not represent exposure in Marte."}, {"provenance": "observed", "statement": "Country-level imports and cargo port calls were lower in January-February than in September-November: dry-bulk imports fell from 1.25 million metric tons in November to 0.96 million in February, total cargo imports from 1.92 million to 1.55 million, and cargo calls from 177-187 to 161. Staple price increased from USD 1.255 per 2.6 kg in November to USD 1.420 in February.", "channels": ["portwatch_import_dry_bulk", "portwatch_import_cargo", "portwatch_portcalls_cargo", "wfp_staple_price"], "event_months": ["2021-11", "2021-12", "2022-01", "2022-02"], "source_url": null, "uncertainty": "National port flows and a country-level price series are indirect proxies; they cannot demonstrate supply or market conditions in Marte."}, {"provenance": "observed", "statement": "Rainfall followed a strong seasonal decline, from 243.9 mm in September to about 3-4 mm in December-January; February rainfall is missing.", "channels": ["chirps_rainfall", "month_of_year"], "event_months": ["2021-09", "2021-10", "2021-11", "2021-12", "2022-01", "2022-02"], "source_url": null, "uncertainty": "The record is country-level and the decline may be seasonal; no production, water-access, or livelihood outcome is observed."}], "rationale": "Retrospectively, the target confirms continued IPC Phase 3 rather than deterioration by June. Before the cutoff, numeric FCS and rCSI values both trended lower overall, while January-February showed elevated national conflict counts, reduced national import/port activity relative to late 2021, and a rising reported staple price. These channels plausibly describe a constrained context that could be consistent with persistence of an adverse IPC phase, but they do not establish causation or local conditions in Marte. The survey-index directions are reported without food-security interpretation because their normalization and severity orientation are undocumented. Seasonal rainfall decline is measured but is not, by itself, evidence of a shock.", "uncertainty": "Most non-survey covariates are country-level whereas the outcome is for Marte; February rainfall is unavailable; and the short series has no counterfactual or local market, displacement, production, or access information. Thus the evidence supports descriptive temporal association only.", "confidence": 0.76}}
|
| 2 |
+
{"sample_id": "000e29c8fa3677df2e22cc48", "dataset_version": "08268e585e0e9028834bf68007a6fb710581b48da0a88ed556fb8502dbfec270", "purpose": "training_supervision_only", "request_sha256": "b790e0918e3be2e69235aefa4e65c10cb8dc04ce0ff769d9eed324a20e863b5f", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENOm2FeSyG11kwrbnA1cKpDwRMavW", "usage": {"completion_tokens": 916, "prompt_tokens": 1976, "total_tokens": 2892, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 1973, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 2, "precursor_patterns": [{"provenance": "observed", "statement": "The observed IPC phase was 2 in February 2020 and remained 2 in June 2020 (phase_change 0).", "channels": ["last_available_phase"], "event_months": ["2020-02", "2020-06"], "source_url": null, "uncertainty": "The supplied phase observations establish no change over this interval, but do not describe conditions within intervening months."}, {"provenance": "observed", "statement": "The normalized rCSI series rose from 0.371 in October 2019 to 0.515 in February 2020, with consecutive increases from November through February. The normalized FCS series fluctuated, peaking at 0.256 in October and ending at 0.237 in February versus 0.178 in September.", "channels": ["rcsi_rt mean", "fcs_rt mean"], "event_months": ["2019-09", "2019-10", "2019-11", "2019-12", "2020-01", "2020-02"], "source_url": null, "uncertainty": "HFID normalization and directionality are undocumented; these are numeric movements only and cannot be translated into food-consumption or coping severity."}, {"provenance": "observed", "statement": "Country-level conflict events increased from 642 in November 2019 to 804 in February 2020; fatalities rose from 925 to 1,627 over the same endpoints after a December low. These data are not specific to Al Ghaydah.", "channels": ["acled_events", "acled_fatalities"], "event_months": ["2019-11", "2019-12", "2020-01", "2020-02"], "source_url": null, "uncertainty": "National conflict counts may not represent exposure in the admin2 area and cannot by themselves establish an effect on its IPC phase."}, {"provenance": "observed", "statement": "Country-level dry-bulk imports declined from 553,054 metric tons in September 2019 to 437,198 in January 2020, then recovered partly to 473,618 in February. Total cargo imports also fell to 688,563 tons in January from 847,737 in December, then reached 713,090 in February; port calls declined from 77 in December to 61 in February.", "channels": ["portwatch_import_dry_bulk", "portwatch_import_cargo", "portwatch_portcalls_cargo"], "event_months": ["2019-09", "2019-12", "2020-01", "2020-02"], "source_url": null, "uncertainty": "The series are national port aggregates; commodity composition, stocks, inland delivery, and relevance to Al Ghaydah are unobserved."}, {"provenance": "observed", "statement": "The reported wheat milling-cost price declined from USD 4.415 per 50 kg in October 2019 to USD 4.01 in December and was unchanged through February 2020.", "channels": ["wfp_staple_price"], "event_months": ["2019-10", "2019-12", "2020-01", "2020-02"], "source_url": null, "uncertainty": "This is a named milling-cost series at country level, not a documented household retail staple-price measure for Al Ghaydah."}], "rationale": "Retrospectively, the supplied target confirms IPC Phase 2 with no phase change by June. Before the target, the local normalized indices showed a rising rCSI number and fluctuating FCS number, but their unknown normalization prevents interpreting either as food-security deterioration or improvement. National conflict activity and fatalities rose into January\u2013February, while imports and port calls were lower than selected late-2019 highs; these are plausible contextual pressures on access and supply but are spatially aggregated and not causal evidence for the local phase outcome. The reported milling-cost price was stable at the end of the observation window. Overall, the observed pattern is compatible with persistence rather than a documented phase transition.", "uncertainty": "Only one pre-target IPC phase and one future phase are supplied. Key channels are country-level rather than admin2-level, February rainfall is missing, and no household income, market availability, displacement, assistance, or local conflict exposure data are provided. Associations across these series should not be treated as causal.", "confidence": 0.82}}
|
| 3 |
+
{"sample_id": "000fb9335ace6c28e73c4a94", "dataset_version": "08268e585e0e9028834bf68007a6fb710581b48da0a88ed556fb8502dbfec270", "purpose": "training_supervision_only", "request_sha256": "ca582b784861e8614458aba6f1e5857a128f0b4b202acb48bbe417aa37f00118", "model": "gpt-5.6-terra", "request_id": "chatcmpl-ENOm22vNb11zBK0rfzF91FbS6ZnaT", "usage": {"completion_tokens": 779, "prompt_tokens": 1953, "total_tokens": 2732, "completion_tokens_details": {"accepted_prediction_tokens": 0, "audio_tokens": 0, "reasoning_tokens": 0, "rejected_prediction_tokens": 0, "text_tokens": null}, "prompt_tokens_details": {"audio_tokens": 0, "cache_write_tokens": 1950, "cached_tokens": 0, "image_tokens": null, "text_tokens": null}}, "annotation": {"observed_future_phase": 1, "precursor_patterns": [{"provenance": "observed", "statement": "IPC phase was 1 in June 2021 and remained phase 1 in the October 2021 target month (phase_change 0).", "channels": ["last_available_phase"], "event_months": ["2021-06", "2021-10"], "source_url": null, "uncertainty": "The supplied target establishes the later phase but does not provide intervening IPC observations."}, {"provenance": "observed", "statement": "The admin2 FCS normalized value rose from 0.090 in January to 0.138 in May, then declined to 0.126 in June. The rCSI normalized value peaked at 0.341 in April and declined to 0.275 by June.", "channels": ["fcs_rt mean", "rcsi_rt mean"], "event_months": ["2021-01", "2021-04", "2021-05", "2021-06"], "source_url": null, "uncertainty": "HFID normalization method and index direction are unknown; these numeric movements cannot be translated into changes in food consumption or coping."}, {"provenance": "observed", "statement": "Country-level cargo imports and dry-bulk imports were variable but reached their six-month highs in June (1.397 million metric tons and 0.968 million metric tons, respectively); cargo port calls stayed within 113\u2013128 per month.", "channels": ["portwatch_import_cargo", "portwatch_import_dry_bulk", "portwatch_portcalls_cargo"], "event_months": ["2021-01", "2021-06"], "source_url": null, "uncertainty": "These are national port measures, not food-specific flows or evidence of access in Saboti."}, {"provenance": "observed", "statement": "Country-level conflict events increased from 11 in April to 31 in May and 32 in June, while reported fatalities increased from 13 in April to 38 in May and 66 in June.", "channels": ["acled_events", "acled_fatalities"], "event_months": ["2021-04", "2021-05", "2021-06"], "source_url": null, "uncertainty": "Country aggregates may not represent conflict exposure in Saboti, and the data do not demonstrate a local food-security effect."}, {"provenance": "observed", "statement": "Recorded country rainfall increased from 20.8 mm in January to 95.4 mm in April and remained 82.1 mm in May; June rainfall is missing.", "channels": ["chirps_rainfall"], "event_months": ["2021-01", "2021-04", "2021-05"], "source_url": null, "uncertainty": "National monthly rainfall does not establish local agricultural production, crop condition, or household food access."}], "rationale": "Retrospectively, the observed outcome is stable IPC phase 1 from June to October. Measured precursor channels show mixed numeric FCS/rCSI movements, rising national import volumes into June, increased national conflict counts and fatalities in May\u2013June, and higher rainfall in April\u2013May than January\u2013March. These patterns do not amount to direct evidence of a causal mechanism for Saboti\u2019s stable phase: the port, conflict, and rainfall series are country-level; staple-price data are entirely unavailable; June rainfall is missing; and normalized FCS/rCSI directions have no documented food-security interpretation. The stable phase is therefore the clearest evidence, with the other channels providing only contextual, non-causal associations.", "uncertainty": "High uncertainty about subnational representativeness and mechanisms. No documented local price, production, market-access, or intervening IPC evidence is supplied, and normalized survey-index direction is unknown.", "confidence": 0.84}}
|
annotations/annotations-terra-pilot-final.manifest.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"requested": 3,
|
| 3 |
+
"valid": 3,
|
| 4 |
+
"budget_exhausted": false,
|
| 5 |
+
"accounted_cost_usd": 0.042918,
|
| 6 |
+
"failures": [],
|
| 7 |
+
"model": "gpt-5.6-terra",
|
| 8 |
+
"conservative_cost_bound_usd": 0.19579400000000002,
|
| 9 |
+
"estimated_usage_usd": 0.042918,
|
| 10 |
+
"pricing_source": "https://developers.openai.com/api/docs/models/gpt-5.6-terra",
|
| 11 |
+
"cached_usage_may_include_previous_runs": true
|
| 12 |
+
}
|
annotations/annotations-terra-pilot.jsonl
ADDED
|
File without changes
|
annotations/annotations-terra-pilot.manifest.json
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"requested": 3,
|
| 3 |
+
"valid": 0,
|
| 4 |
+
"failures": [
|
| 5 |
+
{
|
| 6 |
+
"sample_id": "001e6338a8099dd54cf35585",
|
| 7 |
+
"error": "Annotation references unavailable channel"
|
| 8 |
+
},
|
| 9 |
+
{
|
| 10 |
+
"sample_id": "008741623f7b301f71537290",
|
| 11 |
+
"error": "Annotation references unavailable channel"
|
| 12 |
+
},
|
| 13 |
+
{
|
| 14 |
+
"sample_id": "0055737fcdedc492437f7aea",
|
| 15 |
+
"error": "Annotation references unavailable channel"
|
| 16 |
+
}
|
| 17 |
+
],
|
| 18 |
+
"model": "gpt-5.6-terra",
|
| 19 |
+
"conservative_cost_bound_usd": 0.201606,
|
| 20 |
+
"estimated_usage_usd": 0,
|
| 21 |
+
"pricing_source": "https://developers.openai.com/api/docs/models/gpt-5.6-terra",
|
| 22 |
+
"cached_usage_may_include_previous_runs": true
|
| 23 |
+
}
|