5 ms·
Remarkable how few tokens it needed to get a much better score than other reasoning models. Any chance of contamination?
by causal 2y ago
Remarkable how few tokens it needed to get a much better score than other reasoning models. Any chance of contamination?
- 85392_school 2y agoIt makes me wonder how the token counting was implemented and if it missed the (not sent in API) reasoning.
- freediver 2y agoVaild concern, most likely thinking tokens were not counted due to API reporting changes.