GRAB Benchmark Offers AI Risk Topic Discovery for 10‑K Filings

The GRAB benchmark releases a 1.61 M‑sentence dataset drawn from 8,247 10‑K filings, providing span‑grounded labels for 193 risk terms across 21 fine‑grained types. getnews.me/grab-benchmark-offers-ai... #grabbenchmark #riskdisclosure #10k