Skip to main content
PolicyBench

Kimi K3

Moonshot AI · AI alone, no tools
Current board, snapshot 2026-10-14
United States rank
#8 of 50
US exact
90.6%
United States

United States benchmark

Exact match
90.6%
Bounded score
94.9%
Parse rate
96.8%
Eligibility flags
925/975

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. Self-employment tax97.0%2.1%
  2. Local income tax96.9%0.3%
  3. Person-level WIC eligibility96.6%0.3%
  4. Free school meals eligibility96.0%0.7%
  5. Reduced-price school meals eligibility96.0%0.1%
  6. TANF96.0%0.3%
  7. SSI95.9%2.0%
  8. Person-level Medicaid eligibility94.8%30%
  9. Person-level Early Head Start eligibility94.7%3.1%
  10. Person-level Head Start eligibility94.7%1.2%
  11. Person-level CHIP eligibility93.8%0.2%
  12. Person-level Medicare eligibility93.0%11%
  13. Payroll tax92.7%16%
  14. Federal refundable credits89.0%3.7%
  15. SNAP86.0%4.1%
  16. Federal tax before refundable credits85.7%19%
  17. State refundable credits80.8%0.6%
  18. State tax before refundable credits65.6%6.0%
Table view (exact, within 1%, outputs), hardest first
ProgramExactWithin 1%n
State tax before refundable credits65.6%71.0%93
State refundable credits80.8%80.8%99
Federal tax before refundable credits85.7%89.3%84
SNAP86.0%90.3%93
Federal refundable credits89.0%91.0%100
Payroll tax92.7%93.8%96
Person-level Medicare eligibility93.0%93.0%172
Person-level CHIP eligibility93.8%93.8%177
Person-level Early Head Start eligibility94.7%94.7%38
Person-level Head Start eligibility94.7%94.7%38
Person-level Medicaid eligibility94.8%94.8%173
SSI95.9%95.9%97
Free school meals eligibility96.0%96.0%100
Reduced-price school meals eligibility96.0%96.0%100
TANF96.0%96.0%100
Person-level WIC eligibility96.6%96.6%177
Local income tax96.9%96.9%98
Self-employment tax97.0%97.0%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.