Introducing BigLaw Bench: Global – Harvey
BLB: Global is built to understand models’ ability to deliver on core AI tasks for our customers around the world, starting with the UK, Australia, and Spain.
As we announced earlier this month, the first of our three major extensions to BigLaw Bench (BLB) is our Global dataset. BLB: Global extends the principles of BLB by working with local experts to build benchmarks reflecting the nuance of legal work in global jurisdictions.
Designed in collaboration with our data partners at Mercor, one of the world’s leading companies in the expert data space, BLB: Global is built to understand models’ ability to deliver on the core promises of AI for our customers around the world. This understanding will enable us to ensure that customers across jurisdictions can execute core AI workflows accurately, consistently, and in alignment with local norms.
Our initial BLB: Global dataset more than doubles the size of our public-facing BLB benchmark, adding unique benchmarks for the UK, Australia, and Spain. Throughout the rest of this post, we discuss how we built BLB: Global, what it’s focused on benchmarking, and how we intend to extend it.
To read the article in full, click here.
