GPT-6 Luna
United States benchmark
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.
- Local income tax100.0%0.3%
- Person-level Early Head Start eligibility100.0%3.1%
- Person-level Head Start eligibility100.0%1.2%
- SSI100.0%2.0%
- Person-level WIC eligibility99.4%0.3%
- Self-employment tax99.0%2.1%
- TANF99.0%0.3%
- Person-level CHIP eligibility98.9%0.2%
- Person-level Medicare eligibility98.3%11%
- Free school meals eligibility98.0%0.7%
- Reduced-price school meals eligibility98.0%0.1%
- Person-level Medicaid eligibility97.1%30%
- Federal refundable credits94.9%3.7%
- Payroll tax91.9%16%
- State refundable credits86.7%0.6%
- SNAP86.0%4.1%
- Federal tax before refundable credits77.6%19%
- State tax before refundable credits73.3%6.0%
Table view (exact, within 1%, outputs), hardest first
| Program | Exact | Within 1% | n |
|---|---|---|---|
| State tax before refundable credits | 73.3% | 74.4% | 86 |
| Federal tax before refundable credits | 77.6% | 84.7% | 85 |
| SNAP | 86.0% | 89.2% | 93 |
| State refundable credits | 86.7% | 86.7% | 98 |
| Payroll tax | 91.9% | 94.9% | 99 |
| Federal refundable credits | 94.9% | 96.9% | 98 |
| Person-level Medicaid eligibility | 97.1% | 97.1% | 174 |
| Free school meals eligibility | 98.0% | 98.0% | 100 |
| Reduced-price school meals eligibility | 98.0% | 98.0% | 100 |
| Person-level Medicare eligibility | 98.3% | 98.3% | 172 |
| Person-level CHIP eligibility | 98.9% | 98.9% | 177 |
| Self-employment tax | 99.0% | 99.0% | 100 |
| TANF | 99.0% | 99.0% | 100 |
| Person-level WIC eligibility | 99.4% | 99.4% | 177 |
| Local income tax | 100.0% | 100.0% | 100 |
| Person-level Early Head Start eligibility | 100.0% | 100.0% | 38 |
| Person-level Head Start eligibility | 100.0% | 100.0% | 38 |
| SSI | 100.0% | 100.0% | 97 |
Hardest casesworst misses on positive references
- SNAP116% offReference $2,376 · predicted $5,124Inspect household #079
- State refundable credits106% offReference $2,188 · predicted $4,500Inspect household #032
- State refundable credits100% offReference $25 · predicted $0Inspect household #013
- Federal tax before refundable credits100% offReference $643 · predicted $0Inspect household #023
- SNAP100% offReference $288 · predicted $0Inspect household #027
- State refundable credits100% offReference $40 · predicted $0Inspect household #029
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.