1. What is a potato?
2. What’s the area of a hallway that’s 5m by 1m?
3. sina toki ala toki e toki pona (toki pona for “do you speak toki pona?”. It’s a good litmus for multilingual capabilities and size/variety of training data.
4. If the above goes through: sina sona e toki Tosi anu seme? (“Do you speak German?” in toki pona. I’ve yet to find a model which gets this right.)
5. Generate a bash script to create 10 files, with the names A1.txt through A10.txt
6. Sprichst du Deutsch? (“Do you speak German?” In German.)

It’s a bit random, but it gets it done.

Relevant comic

Frankly, I’d argue LLMs are not the tool for this—not only at a fundamental level (they aren’t the right tool for the job, given hallucination and a host of other factors), but they are also way too resource intensive right now.

Subjectively, how much better do you think this is than Marx-3B, since you’re the creator of both?