logoalt Hacker News

sinuhe69today at 5:09 PM0 repliesview on HN

Recent incident with the Rio 3.5 model clearly shows that many coding models are specifically trained/fine tuned for the benchmarks.