AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
How accurately frontier AI models estimate US tax and benefit amounts, scored against PolicyEngine reference outputs.
Translation pending
Developer toolsPython · TypeScriptGitHub @PolicyEngineOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
LLM benchmark for public policy
Translation pending
Taken from the repository README
Loading the full README. The saved excerpt is shown below.
How well can frontier models calculate tax and benefit outcomes without tools?
PolicyBench measures how well frontier AI models estimate selected household tax
and benefit outputs without tools.
For benchmark scope, snapshot policy, and terminology, see the
benchmark card.
Current public benchmark scenarios are sampled from the certified PolicyEngine
populace US dataset and evaluated under tax year 2026 rules with
PolicyEngine-US.
1. AI alone: Models estimate tax/benefit values using only their training knowledgeRead the full README ↗The full README is unavailable right now. Read it at the original link. The saved excerpt is shown below.
How well can frontier models calculate tax and benefit outcomes without tools?
PolicyBench measures how well frontier AI models estimate selected household tax
and benefit outputs without tools.
For benchmark scope, snapshot policy, and terminology, see the
benchmark card.
Current public benchmark scenarios are sampled from the certified PolicyEngine
populace US dataset and evaluated under tax year 2026 rules with
PolicyEngine-US.
1. AI alone: Models estimate tax/benefit values using only their training knowledgeRead the full README ↗