Llama 4 was released and received a flood of negative reviews within 36 hours. Anonymous employees revealed that they refused to sign the technical report and were even made into emoticons for ridicule. In summary, it is just a "flood of negative reviews". Specifically, people's complaints mainly focus on the code ability. The most straightforward...

36 Hours after the Release of Llama 4, a Flood of Negative Reviews. Anonymous Employees Exposed Refusal to Sign the Technical Report and It Was Made into Emojis for Teasing. In Summary, It's a "Flood of Negative Reviews". Specifically...

via cnBeta.COM Chinese Industry Information Station - Telegram Channel

Telegraph
36 Hours after the Release of Llama 4, a Flood of Negative Reviews. Anonymous Employees Exposed Refusal to Sign the Technical Report and It Was Made into Emojis for Teasing. In Summary...

36 hours after the release of Llama 4, there was a flood of negative reviews. Anonymous employees exposed that they refused to sign the technical report, and it was made into emojis for teasing. In summary, it's a "flood of negative reviews". Specifically, people's complaints mainly focused on code capabilities. The most intuitive one is the classic "ambient programming" ball bounce test, where the ball directly passed through the wall and fell. Reflected in the rankings, the results are quite fragmented. The official evaluation (LiveCodeBench) scores at the time of release and the performance in the large model arena were clearly very good. But in various third - party benchmark tests, the situation mostly reversed directly, ranking at the bottom. It makes people wonder whether the arena ranking is due to data over - fitting or vote - rigging. Just a few days before the release of Llama 4, Joelle, the head of Meta AI research...