Qualification by Calibration: New Benchmark Audits LLM Annotators Like Crowd Workers

windx0303 · x · 2026-09-29

Presented at HCOMP 2026 / CI 2026, the work treats LLM annotators like crowd workers needing qualification. The paper proposes Qualification by Calibration, a readable benchmark for admitting language models to human-computation tasks, adapting crowdsourcing-style worker vetting to LLM-based annotation.

Original post →

More from Research

Research channel →