There were at least two instances of papers that were published during my personal red teaming window where the conclusion was basically language models still can't do X. That sent me, that's still an open question in the broader research community. So I started to think, well, how good is that? All these different benchmarks. And I found exactly what you found also, which is that if you set up your benchmarking script with a slightly dated paradigm, like if you take a 2021 big bench script, for example, and you just run it, first of all, it's set up on like a few shot basis. Second, the structure of that few shot is such that the model is

Get the Snipd
podcast app

Unlock the knowledge in podcasts with the podcast player of the future.
App store bannerPlay store banner

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode

Save any
moment

Hear something you like? Tap your headphones to save it with AI-generated key takeaways

Share
& Export

Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more

AI-powered
podcast player

Listen to all your favourite podcasts with AI-powered features

Discover
highlights

Listen to the best highlights from the podcasts you love and dive into the full episode