←back to thread

Brood War Korean Translations

(blog.sourcedive.net)
314 points todsacerdoti | 1 comments | 17 Jan 25 17:15 UTC | HN request time: 0.28s | source
Show context
nfRfqX5n ◴[17 Jan 25 18:08 UTC] No.42741371[source]▶
Pretty cool and shows a clear issue I’ve seen across any LLM. the language and grammar is so formal/robotic
replies(1): >>42747301 #
1. int_19h ◴[18 Jan 25 10:24 UTC] No.42747301[source]▶
This depends a lot on how the model was fine-tuned. There are many LLaMA tunes that are not at all formal, for example.

With ChatGPT specifically, some forceful prompting can make it sound much better.