3 ms·
Phi3 was trained on benchmarks, it’s contaminated and deceitful. Actual performance is much worse in my experience.
by dannyw 2y ago
Phi3 was trained on benchmarks, it’s contaminated and deceitful. Actual performance is much worse in my experience.
- lostmsu 2y agoPhi-3-Mini has the same ELO on chatbot arena as the oldest GPT-3.5-Turbo. It is an 8GB model (~4B paramters?)