4 ms·
Do you have any examples? The only time I noticed an LLM make a language mistake was when using a quantized model (gemma) with my native language (so much small
by perching_aix 1y ago
Do you have any examples? The only time I noticed an LLM make a language mistake was when using a quantized model (gemma) with my native language (so much smaller training data pool).
- Breza 1y agoNot GP, but I've definitely seen cutting edge LLMs make language mistakes. The most head scratching one I've seen in the past few weeks is when Gemini Pro decided to use <em> and </em> tags to emphasize something that was not code.