Dify — POST /v1/datasets/{tenant2}/retrieve — cross-tenant leakage test (Pub 15 question against corpus without Pub 15 docs)
============================================================================================================================

{
  "query": {
    "content": "What is the Social Security wage base for 2026 per IRS Publication 15?"
  },
  "records": [
    {
      "segment": {
        "id": "b7c4a633-7363-402b-b5cb-291bc633f9a0",
        "position": 23,
        "document_id": "7adb4ab2-e40b-49f2-b2a6-7caa094bf92c",
        "content": "Recent Economic and Financial Developments 15\r\nmeasures discussed here remain positive even while the aggregate real wage growth measures discussed elsewhere in this report have turned negative .)\r\nThe recent cooling in real wage growth is visible across all broad demographic groups, including \namong some groups for whom, up until a year ago, wage growth had remained robust . Wage growth \nfor nonwhite workers had been a bit stronger than for white workers from 2022 through mid-2024 \nbut has been similar to white workers’ since then (fi gure C, top-right panel) . Similarly, wage growth for \nworkers with a high school diploma or less was strong relative to other groups in the post-pandemic \ntight labor market; however, as labor market conditions softened in 2024, wage growth for this \ngroup fell below that for college-educated workers in early 2024 and has been slowing, on net, since \nmid-2024 (fi gure C, bottom-left panel) . Some of this softening may be spreading to workers with an \nassociate’s degree, for whom real wage growth has eased more signifi cantly in the past year and currently stands below that for other education groups . Finally, wages for men and women largely grew \nin tandem until the middle of 2024, but real wage growth for women was a bit stronger than that for \nmen from mid-2024 through mid-2025 (fi gure C, bottom-right panel) . However, in recent months, wage \ngrowth for women has tapered a bit and is now a touch below that of men .\r\nBox 1—continued\r\nFigure C. Median real wage growth, by group\r\nMonthly\r\n−4\r\n−3\r\n−2\r\n−1\r\n0\r\n1\r\n2\r\n3\r\n4\r\n2020 2022 2024 2026\r\n1st quartile\r\n2nd quartile\r\n3rd quartile\r\n4th quartile\r\nPercent\r\nWage quartiles\r\nMonthly\r\n−4\r\n−3\r\n−2\r\n−1\r\n0\r\n1\r\n2\r\n3\r\n4\r\n2020 2022 2024 2026\r\nNonwhite\r\nWhite\r\nPercent\r\nRace\r\nMonthly\r\n−4\r\n−3\r\n−2\r\n−1\r\n0\r\n1\r\n2\r\n3\r\n4\r\n2020 2022 2024 2026\r\nHigh school or less\r\nAssociate’s degree\r\nBachelor’s degree or more\r\nPercent\r\nEducational attainment\r\nMonthly\r\n−4\r\n−3\r\n−2\r\n−1\r\n0\r\n1\r\n2\r\n3\r\n4\r\n2020 2022 2024 2026\r\nWomen\r\nMen\r\nPercent\r\nSex\r\nNote: Series show 12-month moving averages of the median percent change in the hourly wage of individuals \nobserved 12 months apart, deflated by the 12-month moving average of the 12-month percent change in the \npersonal consumption expenditures price index . In the top-left panel, workers are assigned to wage quartiles \nbased on the average of their wage reports in both Current Population Survey outgoing rotation group \ninterviews; workers in the lowest 25 percent of the average wage distribution are assigned to the 1st quartile, \nand those in the top 25 percent are assigned to the 4th quartile .",
        "sign_content": null,
        "answer": null,
        "word_count": 2627,
        "tokens": 682,
        "keywords": [],
        "index_node_id": "1f5a09f0-156b-4acd-aeef-483cc9843d32",
        "index_node_hash": "27d06a2c53bc38fcee727702b7fd4ad035b2ac148d40ab9a184429cff64f4ebd",
        "hit_count": 0,
        "enabled": true,
        "disabled_at": null,
        "disabled_by": null,
        "status": "completed",
        "created_by": "8426f7fc-70c4-4cc8-a4f5-c2f9575c063a",
        "created_at": 1787780884,
        "indexing_at": 1787780887,
        "completed_at": 1787780891,
        "error": null,
        "stopped_at": null,
        "document": {
          "id": "7adb4ab2-e40b-49f2-b2a6-7caa094bf92c",
          "data_source_type": "upload_file",
          "name": "03_longform_fed-monetary-policy-report-2026-07.pdf",
          "doc_type": null,
          "doc_metadata": null
        }
      },
      "child_chunks": [],
      "score": 0.38579583,
      "tsne_position": null,
      "files": [],
      "summary": null
    },
    {
      "segment": {
        "id": "7ac076a0-1635-4d7c-9112-48cdc4743114",
        "position": 26,
        "document_id": "7adb4ab2-e40b-49f2-b2a6-7caa094bf92c",
        "content": "18 Monetary Policy Report\r\nA task force will explore productivity and jobs\r\nOne of the recently announced independent task forces will explore the subject of productivity \nand jobs. It will survey the pace, the reach, and the economic effects of generative AI—a new \ngeneral-purpose technology—and other innovations. The task force will explore the implications \nfor the Federal Reserve in pursuit of its maximum-employment and price-stability mandates.\r\nWage growth remains solid and, with persistently strong productivity growth, is \nroughly consistent with 2 percent inflation over time\r\nMeasures of nominal wage growth remain \nsolid, although they have edged down this \nyear on net. Total hourly compensation \nfor private-sector workers, as measured \nby the employment cost index, increased \n3.4 percent over the year ending in March, \nunchanged from its pace a year earlier but \nwell below its peak increase of 5.5 percent in \nmid-2022 (figure 17). However, other measures of labor compensation growth, such \nas average hourly earnings (a less comprehensive measure of compensation) and the \nFederal Reserve Bank of Atlanta’s Wage \nGrowth Tracker (which reports the median \n12-month wage growth of individuals responding to the Current Population Survey), have \nmoved lower over the past year.\r\nAlthough wage growth remains a touch above its pre-pandemic pace, persistently strong labor \nproductivity growth suggests that current nominal wage growth is roughly consistent with \n2 percent inflation over time.\r\nSolid nominal wage gains have nonetheless been outpaced by overall price inflation over the past \nyear. With the recent jump in energy prices boosting the change in PCE prices to 4.1 percent over \nthe 12 months ending in May, the purchasing power of workers’ wages, in the aggregate, declined \nsomewhat over this period. That said, for the year ending in May 2025, wage gains outpaced \nprice inflation. The effects on individual households, though, depend in part on workers’ circumstances—because nominal wage changes vary significantly across industry and occupation \nand because households consume different baskets of goods than the one represented in the \naggregate PCE price index. (For details on how real wage gains have differed across demographic \ngroups, see the box “Employment and Earnings across Demographic Groups.”)\r\nFigure 17. Measures of change in hourly \ncompensation\r\n0\r\n1\r\n2\r\n3\r\n4\r\n5\r\n6\r\n7\r\n8\r\n9\r\n10\r\n11\r\n2016 2018 2020 2022 2024 2026\r\nAtlanta Fed’s Wage Growth Tracker\r\nAverage hourly earnings, private sector\r\nEmployment cost index, private sector\r\nPercent change from year earlier\r\nNote: For the Atlanta Fed’s Wage Growth Tracker, \nthe data are shown as a 3-month moving average \nof the 12-month percent change; for private-sector \naverage hourly earnings, the data are 12-month \npercent changes and extend through June 2026; for \nthe private-sector employment cost index, change is \nover the 12 months ending in the last month of each \nquarter.",
        "sign_content": null,
        "answer": null,
        "word_count": 2984,
        "tokens": 640,
        "keywords": [],
        "index_node_id": "02388fad-dbbd-459a-8f36-5665ef781efd",
        "index_node_hash": "d55ff73551230e46068289ac51c8199b0be85fc445e0fc8f3bb300afc981fad8",
        "hit_count": 0,
        "enabled": true,
        "disabled_at": null,
        "disabled_by": null,
        "status": "completed",
        "created_by": "8426f7fc-70c4-4cc8-a4f5-c2f9575c063a",
        "created_at": 1787780884,
        "indexing_at": 1787780887,
        "completed_at": 1787780891,
        "error": null,
        "stopped_at": null,
        "document": {
          "id": "7adb4ab2-e40b-49f2-b2a6-7caa094bf92c",
          "data_source_type": "upload_file",
          "name": "03_longform_fed-monetary-policy-report-2026-07.pdf",
          "doc_type": null,
          "doc_metadata": null
        }
      },
      "child_chunks": [],
      "score": 0.36138237,
      "tsne_position": null,
      "files": [],
      "summary": null
    },
    {
      "segment": {
        "id": "6889530b-3e67-4bc4-9a87-0d367a7934e5",
        "position": 18,
        "document_id": "7adb4ab2-e40b-49f2-b2a6-7caa094bf92c",
        "content": "12 Monetary Policy Report\r\nBox 1. Employment and Earnings across Demographic Groups\r\nEmployment disparities across sex, race, ethnicity, and education groups—some of which reached \nrecord lows in 2023 and early 2024 in an especially tight labor market—continue to be relatively \nnarrow compared with historical levels . However, despite the labor market progress for many demographic groups in recent years, signifi cant disparities in absolute levels across groups remain . Additionally, the robust real wage gains experienced by some historically disadvantaged groups in recent \nyears have since moderated as labor market tightness has eased and consumer price infl ation has \nremained elevated .\r\nAmong prime-age people (aged 25 to 54), the employment-to-population (EPOP) ratio for Black or \nAfrican American workers softened considerably in the fi rst half of 2025 but has since partially recovered and, in recent months, has been a touch above its average level in 2019 (fi gure A, left panel) .1\r\nThis development primarily refl ects an increase in the unemployment rate for this group over early \n2025 that has since partially reversed .2 At the same time, the employment rate for white workers has \ncontinued to gradually drift higher over the past year . As a result, the EPOP ratio gap between Black \nand white individuals has widened, on net, over the past year, although it remains narrow compared \nwith its historical level .3 By contrast, the employment rate for Hispanic or Latino workers has held \nroughly steady, on net, over the past year, while the employment rate for Asian workers has edged \n1 Because of the government shutdown in 2025, the Bureau of Labor Statistics did not collect October employment and wage \ndata . Therefore, October 2025 employment and nominal wage values are estimated as the average of September 2025 and \nNovember 2025 values .\r\nThe missed collection of employment data for October 2025 may complicate the comparison of employment and labor force \nparticipation rates across demographic groups because of the varying effect across groups of a form of survey sampling error \nreferred to as “excess rotation group” bias . For a discussion of this possible bias, see Hie Joo Ahn and James D . Hamilton \n(2022), “Measuring Labor-Force Participation and the Incidence and Duration of Unemployment,” Review of Economic \nDynamics, vol . 44 (April), pp . 1–32 .\r\n2 The EPOP ratio—that is, employment divided by population—can also be expressed as LFPR * (1 – UR), where LFPR is the \nlabor force participation rate and UR is the unemployment rate . The EPOP ratio therefore decreases as the LFPR decreases or \nas the unemployment rate increases . The EPOP ratio is multiplied by 100 for presentation purposes in the figures .\r\n3 In figures A and B, EPOP ratios are shown indexed to their 2019 average; therefore, gaps between groups are not readily \nevident . Through December 2025, EPOP ratios by sex and education are constructed using harmonized microdata weights . \nFor more information, see John Coglianese, Seth Murray, and Christopher J . Nekarda (2025), “Harmonized Population and \nLabor Force Statistics,” Finance and Economics Discussion Series 2025-057 (Washington: Board of Governors of the \nFederal Reserve System, August), https://doi .org/10 .17016/FEDS .2025 .057 .\r\n(continued)\r\nFigure A. Prime-age employment-to-population ratios compared with the 2019 average ratio,\r\nby group\r\nMonthly\r\n−14\r\n−12\r\n−10\r\n−8\r\n−6\r\n−4\r\n−2\r\n0\r\n2\r\n4\r\n2020 2022 2024 2026\r\nAsian\r\nWhite\r\nBlack or African American\r\nHispanic or Latino\r\nPercentage points\r\nRace and ethnicity\r\nMonthly\r\n−14\r\n−12\r\n−10\r\n−8\r\n−6\r\n−4\r\n−2\r\n0\r\n2\r\n4\r\n2020 2022 2024 2026\r\nWomen, some college or more\r\nWomen, high school or less\r\nMen, some college or more\r\nMen, high school or less\r\nPercentage points\r\nSex and educational attainment\r\nNote: The data are 3-month moving averages . The data by race and ethnicity extend through June 2026 . \nPrime age is 25 to 54 .",
        "sign_content": null,
        "answer": null,
        "word_count": 3965,
        "tokens": 918,
        "keywords": [],
        "index_node_id": "761de35b-5d47-49ea-875a-2ebbcbbe0eb1",
        "index_node_hash": "aae74ba7e4c9f8f1ae3bb3441cf8ce903aa8a3fbb7404019e1c3bac2046445d2",
        "hit_count": 0,
        "enabled": true,
        "disabled_at": null,
        "disabled_by": null,
        "status": "completed",
        "created_by": "8426f7fc-70c4-4cc8-a4f5-c2f9575c063a",
        "created_at": 1787780884,
        "indexing_at": 1787780887,
        "completed_at": 1787780891,
        "error": null,
        "stopped_at": null,
        "document": {
          "id": "7adb4ab2-e40b-49f2-b2a6-7caa094bf92c",
          "data_source_type": "upload_file",
          "name": "03_longform_fed-monetary-policy-report-2026-07.pdf",
          "doc_type": null,
          "doc_metadata": null
        }
      },
      "child_chunks": [],
      "score": 0.3506652,
      "tsne_position": null,
      "files": [],
      "summary": null
    }
  ]
}