3 ms·

I have a question, as it happens: Do you think the benchmarks and models were trained on benchmark datasets to skew the results, even though in real-world appli

I have a question, as it happens: Do you think the benchmarks and models were trained on benchmark datasets to skew the results, even though in real-world applications we realize they're not that great?

sinuhe69 4mo ago
Recent incident with the Rio 3.5 model clearly shows that many coding models are specifically trained/fine tuned for the benchmarks.
- RDTvlokip 4mo ago
  That's what I thought