Skip to main content
PolicyBench

GPT-6 Sol

OpenAI · AI alone, no tools
Current board, snapshot 2026-09-22
United States rank
#1 of 42
US exact
94.2%
United States

United States benchmark

Exact match
94.2%
Bounded score
97.8%
Parse rate
100.0%
Eligibility flags
969/976

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. Free school meals eligibility100.0%0.7%
  2. Local income tax100.0%0.3%
  3. Person-level Early Head Start eligibility100.0%3.1%
  4. Person-level Head Start eligibility100.0%1.2%
  5. Person-level Medicare eligibility100.0%11%
  6. Person-level WIC eligibility100.0%0.3%
  7. Self-employment tax100.0%2.1%
  8. Person-level CHIP eligibility99.4%0.2%
  9. Reduced-price school meals eligibility99.0%0.1%
  10. TANF99.0%0.3%
  11. Payroll tax99.0%16%
  12. SSI99.0%2.0%
  13. Person-level Medicaid eligibility97.1%30%
  14. Federal refundable credits94.9%3.7%
  15. SNAP89.2%4.1%
  16. State refundable credits85.7%0.6%
  17. Federal tax before refundable credits84.7%19%
  18. State tax before refundable credits74.4%6.0%
Table view (exact, within 1%, outputs), hardest first
ProgramExactWithin 1%n
State tax before refundable credits74.4%76.7%86
Federal tax before refundable credits84.7%89.4%85
State refundable credits85.7%85.7%98
SNAP89.2%90.3%93
Federal refundable credits94.9%95.9%98
Person-level Medicaid eligibility97.1%97.1%174
SSI99.0%99.0%97
Payroll tax99.0%99.0%99
Reduced-price school meals eligibility99.0%99.0%100
TANF99.0%99.0%100
Person-level CHIP eligibility99.4%99.4%177
Free school meals eligibility100.0%100.0%100
Local income tax100.0%100.0%100
Person-level Early Head Start eligibility100.0%100.0%38
Person-level Head Start eligibility100.0%100.0%38
Person-level Medicare eligibility100.0%100.0%172
Person-level WIC eligibility100.0%100.0%177
Self-employment tax100.0%100.0%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 7,583 rows whose legacy threshold score is below 1. That universe contains 7,579 of 7,579 exact-match misses and 4 exact hits. Another 1,843 rows with bounded score below 100 were not selected and have no audit annotation.