Your Model's IQ Won't Impress LMSys
·48 words·1 min
A valuable lesson for folks developing large models:
A model with much better reasoning skills does not mean it will rank higher on LMSys!
But why?
- Reasoning is actually very small part of queries
- Regular folks aren’t asking a ton of clever puzzles and MMLU style… https://x.com/lmarena_ai/status/1788363018449166415