NIAT-Pro is a benchmark for evaluating how well large language models understand and reason over large tables under controlled variations of tabular format, table size, and information position.
It extends the original Needle-In-A-Table setting from simple cell lookup to one-hop, two-hop, and four-hop tasks, and studies performance across 11 table representations:
NIAT-Pro is designed… See the full description on the dataset page:
https://huggingface.co/datasets/NIAT-Pro/needle-in-a-table-pro.
