←back to thread

Teuken-7B-Base and Teuken-7B-Instruct: Towards European LLMs (2024)

(arxiv.org)
246 points doener | 4 comments | 15 Apr 25 10:17 UTC | HN request time: 1.807s | source
Show context
ozgune ◴[15 Apr 25 12:12 UTC] No.43691597[source]▶
I had a related, but orthogonal question about multilingual LLMs.

When I ask smaller models a question in English, the model does well. When I ask the same model a question in Turkish, the answer is mediocre. When I ask the model to translate my question into English, get the answer, and translate the answer back to Turkish, the model again does well.

For example, I tried the above with Llama 3.3 70B, and asked it to plan me a 3-day trip to Istanbul. When I asked Llama to do the translations between English <> Turkish, the answer was notably better.

Anyone else observed a similar behavior?

replies(11): >>43691620 #>>43691751 #>>43691774 #>>43692427 #>>43692596 #>>43692803 #>>43692874 #>>43693906 #>>43695475 #>>43698229 #>>43698667 #
laurent_du ◴[15 Apr 25 13:43 UTC] No.43692596[source]▶
ChatGPT is very informal and talks like a millennial when I ask questions in French. I hate it.
replies(5): >>43692843 #>>43692880 #>>43693378 #>>43694306 #>>43697616 #
1. dingnuts ◴[15 Apr 25 14:03 UTC] No.43692843[source]▶
sorry u hate a whole generation
replies(1): >>43693172 #
2. mdp2021 ◴[15 Apr 25 14:22 UTC] No.43693172[source]▶
That's not a "generation", that is a "portrait" (a characterization).
replies(1): >>43698929 #
3. Der_Einzige ◴[15 Apr 25 21:59 UTC] No.43698929[source]▶
The most french response of all... "Euh, en fait.."
replies(1): >>43702922 #
4. mdp2021 ◴[16 Apr 25 08:27 UTC] No.43702922{3}[source]▶
Britons have "actually", and German speakers spawned the great wave of analytics in the past century (that Britons dominated).