10h ago

Expert Questions AI Benchmark Relevance For Real-World Science Applications

——0——
Original post
OPAnshul Kundaje#1675Anshul Kundaje@ANSHULKUNDAJE

There are the various benchmarks but unclear to me how representative they are for real life use cases in science. Again, all very nebulous from a users perspective especially given the commercial closed source nature of most of the tools. 5/5

12:01 AM · May 21, 2026 View on X
REPLYAnshul Kundaje#1675Anshul Kundaje@ANSHULKUNDAJE

@andrewwhite01 Not sure I understand. Can you expand.

Andrew White 🐦‍⬛Andrew White 🐦‍⬛@andrewwhite01

@anshulkundaje The 9% is pretty misleading. This is an extremely well-studied problem and getting up to the September 5th 2025 record seems pretty good.

3:13 PM · May 21, 2026 · 215 Views
3:48 PM · May 21, 2026 · 93 Views
Expert Questions AI Benchmark Relevance For Real-World Science Applications · Digg