Claude Sonnet 5
United States benchmark
Serving sensitivity This row ran under the board's forced answer-tool call, which switches Claude's extended thinking off, one output per request. The same model, re-run with tool_choice: auto and the whole household in one request, scores 86.1% exact (+12.3) and would rank #20 on the live board; the board keeps the request shape its model card records. Read the note.
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.
- Person-level Early Head Start eligibility100.0%3.1%
- Reduced-price school meals eligibility100.0%0.1%
- Free school meals eligibility99.0%0.7%
- Local income tax99.0%0.3%
- TANF99.0%0.3%
- Person-level Medicare eligibility98.8%11%
- SSI97.9%2.0%
- Self-employment tax96.0%2.1%
- Person-level Head Start eligibility94.7%1.2%
- Person-level WIC eligibility94.4%0.3%
- Person-level CHIP eligibility91.5%0.2%
- State refundable credits80.6%0.6%
- Payroll tax80.0%16%
- Federal refundable credits79.6%3.7%
- SNAP74.2%4.1%
- Person-level Medicaid eligibility74.0%30%
- State tax before refundable credits62.4%6.0%
- Federal tax before refundable credits51.9%19%
Table view (exact, within 1%, outputs), hardest first
| Program | Exact | Within 1% | n |
|---|---|---|---|
| Federal tax before refundable credits | 51.9% | 53.2% | 79 |
| State tax before refundable credits | 62.4% | 63.5% | 85 |
| Person-level Medicaid eligibility | 74.0% | 74.0% | 173 |
| SNAP | 74.2% | 74.2% | 93 |
| Federal refundable credits | 79.6% | 79.6% | 98 |
| Payroll tax | 80.0% | 90.5% | 95 |
| State refundable credits | 80.6% | 81.6% | 98 |
| Person-level CHIP eligibility | 91.5% | 91.5% | 177 |
| Person-level WIC eligibility | 94.4% | 94.4% | 177 |
| Person-level Head Start eligibility | 94.7% | 94.7% | 38 |
| Self-employment tax | 96.0% | 96.0% | 100 |
| SSI | 97.9% | 97.9% | 97 |
| Person-level Medicare eligibility | 98.8% | 98.8% | 172 |
| Free school meals eligibility | 99.0% | 99.0% | 100 |
| Local income tax | 99.0% | 99.0% | 100 |
| TANF | 99.0% | 99.0% | 100 |
| Person-level Early Head Start eligibility | 100.0% | 100.0% | 38 |
| Reduced-price school meals eligibility | 100.0% | 100.0% | 100 |
Hardest casesworst misses on positive references
- State refundable credits6216% offReference $19 · predicted $1,200Inspect household #043
- Federal tax before refundable credits2341% offReference $500 · predicted $12,200Inspect household #046
- SNAP1358% offReference $288 · predicted $4,200Inspect household #030
- SNAP583% offReference $288 · predicted $1,968Inspect household #108
- Federal refundable credits488% offReference $40 · predicted $234Inspect household #066
- State refundable credits200% offReference $95 · predicted $285Inspect household #023
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 7,527 rows whose legacy threshold score is below 1. That universe contains 7,523 of 7,523 exact-match misses and 4 exact hits. Another 2,072 rows with bounded score below 100 were not selected and have no audit annotation.