Jack Herrera reported that over 15 million people depend on the Rio Grande for their drinking water. ^[1] Jack Herrera, a Texas Monthly contributor, stated that less than 15% of the natural flow of the Rio Grande reaches the delta. ^[2] Researchers found that all three evaluated systems exhibited significant degradation on out-of-distribution inputs. ^[3] Jack Herrera reported that 87% of the water from reservoirs along the river goes to agriculture. ^[4] Jack Herrera reported that the Rio Grande runs dry for hundreds of miles, a stretch locals call 'the forgotten reach' south of El Paso. ^[5] Encoder models primarily failed on unseen surface forms and boundary detection. ^[6] Generative LLMs exhibited entity-type confusion and generation instability. ^[7] Researchers constructed a stress test benchmark spanning seven categories of natural distribution shift to evaluate PII detection systems. ^[8]