Remix.run Logo
simianwords 3 days ago

There’s no way this is right. I checked complicated ones with the latest thinking model. Can someone come up with a counter example?

Edit: here’s what I tried https://chatgpt.com/share/69cebb52-56a8-838f-969c-c47308262a...

stratos123 3 days ago | parent | next [-]

Did you use the exact API call shown in the paper? I am unable to replicate the paper's counterexamples via the chat UI, but that's not very surprising (if the LLM already only fails a few cases out of thousands, the small differences in context between API and chat might fix them).

simianwords 3 days ago | parent [-]

I tried this https://chatgpt.com/share/69cebb52-56a8-838f-969c-c47308262a...

pton_xd 3 days ago | parent | prev [-]

"in this paper we primarily evaluate the LLM itself without external tool calls."

Maybe this is a factor?

simianwords 3 days ago | parent [-]

No tools were used.

chromacity 3 days ago | parent [-]

IIRC, web chat often uses tools / code without surfacing this information in any obvious way.