6 ms·

I think it's most illustrative to see the sample battles (H2H) that LMArena released [1]. The outputs of Meta's model is too verbose and too 'yappy' IMO. And lo

by ekojs 1y ago

I think it's most illustrative to see the sample battles (H2H) that LMArena released [1]. The outputs of Meta's model is too verbose and too 'yappy' IMO. And looking at the verdicts, it's no wonder by people are discounting LMArena rankings. [1]: https://huggingface.co/spaces/lmarena-ai/Llama-4-Maverick-03-26-Experimental_battles https://huggingface.co/spaces/lmarena-ai/Llama-4-Maverick-03...

smeeth 1y ago
In fairness, 4o was like this until very recently. I suspect it comes from training on COT data from larger models.
ed 1y ago
Yep, it’s clear that many wins are due to Llama 4’s lowered refusal rate which is an effective form of elo hacking.