OpenAI说,ChatGPT现在最好遵守指示以避免破碎,但结果却不一致。 OpenAI says ChatGPT now better follows instructions to avoid em dashes, but results are inconsistent.

OpenAI CEO Sam Altman宣布,ChatGPT现在可靠地遵循用户指示,避免使用冲浪,他称之为“小但幸福的胜利”。 OpenAI CEO Sam Altman announced that ChatGPT now reliably follows user instructions to avoid using em dashes, a change he called a "small but happy win." 更新的目的是减少AI可探测性和改进书面控制,结果好坏参半,因为一些用户报告说,尽管命令明确,但该模型仍然使用反射线。 The update, aimed at reducing AI detectability and improving writing control, was met with mixed results, as some users reported the model still used em dashes despite clear commands. 如何使该特征得以实现的问题依然存在,各种互动的绩效各不相同。 Confusion over how to enable the feature persisted, and performance varied across interactions. 这一宣布恰逢 " 团体聊天 " (Group Chats)这一新的协作特点的有限推出,以及OpenAI扩大产品(包括Sora和Atlas)方面继续取得进展。 The announcement coincided with the limited rollout of Group Chats, a new collaborative feature, and continued progress in OpenAI’s broader product expansion, including Sora and Atlas.