Chatgpt’in alıntıları nasıl ele aldığını araştırıyordum ve görünüşe göre, yorumunuzun en iyi derecelendirilmiş olup olmadığı veya% 1 yorumcunuz olup olmadığınızı umursamıyor. Chatgpt, karma ve popülerlik oynamak kolay olduğu için yorumlara hiç güvenmiyor. Sadece Wikipedia, ARXIV, yayınlanmış makaleler ve resmi yayınlar gibi doğrulanmış kaynaklardan gelen bilgileri gerçekten ele alıyor.
Burada gerçek nedir?
LLMs aren’t looking at things like links, karma, etc. It’s more about consensus.
If it sees the majority of documents contain a specific point of view on a topic, it takes that point of view as a source of “truth” and may cite one or more of those documents, including Reddit comments.
It will also often include other points of view in its answers and output if it has found a fair number of documents citing an alternate view or answer in its training corpus.
For things like AI Mode and ChatGPT Search, it is a lot more about what it finds in top ranking documents (in Google or Bing, respectively), since that is what it is relying on for answers as much as or more than its training documents.
AI Mode will cite Reddit threads and comments fairly often because they show up so much in Google Search.
It depends on the country and topic.
In the United States, for many of our clients, Reddit is a top 5 source of citations. Sometimes number 1.
For other clients, especially those outside the US, reddit never shows up at all.