DELTA Water Lab @ PolyU – Data-driven Environmental Laboratory for Technology and Analytics

We study how water systems can use less carbon, recover more resources and put AI to responsible use—all while fulfilling their essential purpose: providing clean and safe water for all.

Regulatory limits from the U.S., China, the EU and WHO drawn as rows around their true values. Answers recalled from model memory land off the limit, most often outside the U.S.; retrieving the regulation text pulls nearly all of them back.

How often does AI hallucinate on environmental regulatory facts?

Open-weight language models got 50–70% of 505 water, air and waste regulatory limits wrong, most often for non-U.S. standards. Looking inside the models, we traced recall of these limits to early-to-middle layers—but editing one limit there often shifted others in the same standard. Grounding answers in the regulation text proved the safer fix.

Scattered points gather into a connected water droplet, suggesting environmental evidence coming together.

Putting language models into environmental context

What does AI need to work reliably on environmental problems? Our perspective connects environmental data, traceable evidence and tools—with physical constraints and human oversight throughout.

Read paper

River tributaries carry treated wastewater toward a downstream drinking-water intake. — A drinking water plant's upstream connectionsDRINCS 2.0 network excerpt

What reaches our drinking water?

Our work on de facto reuse traces how treated wastewater contributes to surface drinking-water supplies.

Read paper Explore the research map