active
Human-Centered Table Unionability
Current PhD research on human and large-language-model judgments of table unionability, metacognitive signals, label calibration, and human–AI combinations for data discovery and integration.
This work studies what human and model judgments reveal about whether tables can be meaningfully combined. It examines confidence, decision time, and interaction traces as additional evidence for evaluating labels and supporting union search.
The TUNE benchmark data is available on GitHub.
- Table unionability
- Human-centered data discovery
- Large language models
- Metacognitive signals