logoalt Hacker News

grendelt • today at 3:50 AM • 4 replies • view on HN

It's great fun asking it questions like "who is the Dalai Lama?", "where is TSMC headquartered?", "why do people post Winnie the Pooh in reference to Xi?", "who controls the South China Sea?", "why was Chiang Kai Shek fighting Mao?", "why were students in Tiananmen Square?"

It'll gush about how to make moon cakes and where the tallest and longest bridges are and how to run qwen on a raspberry pi - but it can't seem to manage some basic Wikipedia questions.


Replies

lanyard-textile • today at 5:04 AM

I suspect this happens when you run the inference in China or first party from the model maker. Out of all the open weight models I've used on US located hardware, they're fine answering these.

Anecdotal though, N = 1.

robertkarl • today at 3:53 AM

I've done some work on this and a lot of the refusals are in the API. The weights are more permissive at these types of things. Try it again when we get some of the smaller Qwen 4 weights.

dpkirchner • today at 4:07 AM

Qwen3.8 flash next handles these questions just fine, what should I expect to see? It correctly reports who the Dalai Lama is, that TSLC is in Taiwan (Hsinchu), explains why people post Winnie the Pooh memes, etc.

rolymath • today at 4:41 AM

Chinese history is closed-weight. Chinese trade is open-weight.

American history is open-weight. American trade is closed-weight.