On: DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL

[whimsicalism]

google absolutely games for lmsys benchmarks with markdown styling. r1 is better than google flash thinking, you are putting way too much faith in lmsys