Skip to main content
PolicyBench

GLM-5.2

Z.ai · AI alone, no tools
Current board, snapshot 2026-10-14
United States rank
#42 of 50
US exact
76.6%
United States

United States benchmark

Exact match
76.6%
Bounded score
86.2%
Parse rate
93.0%
Eligibility flags
892/975

Exact match by programall 18 output groups, best first

Bar length is the unweighted exact-match rate on that program's outputs; the right column is the program's share of the headline weight, which follows household dollars, so the bars do not average to the headline.

  1. Reduced-price school meals eligibility94.0%0.1%
  2. TANF94.0%0.3%
  3. Local income tax93.9%0.3%
  4. Person-level CHIP eligibility93.2%0.2%
  5. Free school meals eligibility93.0%0.7%
  6. SSI92.8%2.0%
  7. Person-level WIC eligibility92.7%0.3%
  8. Person-level Medicare eligibility91.3%11%
  9. Self-employment tax91.0%2.1%
  10. Person-level Early Head Start eligibility89.5%3.1%
  11. Person-level Head Start eligibility89.5%1.2%
  12. Person-level Medicaid eligibility87.3%30%
  13. SNAP81.7%4.1%
  14. Federal refundable credits79.0%3.7%
  15. Payroll tax75.0%16%
  16. State refundable credits74.7%0.6%
  17. State tax before refundable credits49.5%6.0%
  18. Federal tax before refundable credits47.6%19%
Table view (exact, within 1%, outputs), hardest first
ProgramExactWithin 1%n
Federal tax before refundable credits47.6%47.6%84
State tax before refundable credits49.5%51.6%93
State refundable credits74.7%74.7%99
Payroll tax75.0%80.2%96
Federal refundable credits79.0%79.0%100
SNAP81.7%82.8%93
Person-level Medicaid eligibility87.3%87.3%173
Person-level Early Head Start eligibility89.5%89.5%38
Person-level Head Start eligibility89.5%89.5%38
Self-employment tax91.0%91.0%100
Person-level Medicare eligibility91.3%91.3%172
Person-level WIC eligibility92.7%92.7%177
SSI92.8%92.8%97
Free school meals eligibility93.0%93.0%100
Person-level CHIP eligibility93.2%93.2%177
Local income tax93.9%93.9%98
Reduced-price school meals eligibility94.0%94.0%100
TANF94.0%94.0%100

Hardest casesworst misses on positive references

Scores are from the frozen manuscript snapshot under the AI-alone condition: structured responses over the same household facts (whole-scenario or in output subsets, per the serving-configuration table), no tools, graded against PolicyEngine reference outputs. See the leaderboard and paper for methodology. The frozen US annotations cover 8,765 rows whose legacy threshold score is below 1. That universe contains 8,761 of 8,761 exact-match misses and 4 exact hits. Another 2,329 rows with bounded score below 100 were not selected and have no audit annotation.