Evidence item

The counted passage establishes the archive composition and US setting but omits the LLM-prediction evaluation purpose and 2026 timing. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 25423

Source: Large language models can predict the results of social science experiments · locator {} · excerpt hash sha256:3a95238e0cedb21106b8120d74630839a44b2c9edd25cdc5f6cbe5574b66ecf7

Here we built an archive of 70 preregistered, nationally representative survey experiments in the USA involving 469 experimental effects and 119,330 participants.