不得不承认，Google 团队在训练 Nano Banana 的路线上，采用「文字渲染」作为一项重要指标，是很成功的选择。之前团队在技术分享播客上就曾提到，如果能把图片上的文字准确地生成，那么整个图片生成也会没有太大的问题。
"The stories and experiences Neil has been able to share with us are insane."
，这一点在旺商聊官方下载中也有详细论述
Random-generator or custom URL slugs
Two subtle ways agents can implicitly negatively affect the benchmark results but wouldn’t be considered cheating/gaming it are a) implementing a form of caching so the benchmark tests are not independent and b) launching benchmarks in parallel on the same system. I eventually added AGENTS.md rules to ideally prevent both. ↩︎
I also put "-" between things that look like multiple words, but would be written as one word today.