4 ms·
You build an LLM that makes mistakes. Then you build another LLM to find the mistakes the first LLM makes, and, surprise: > CriticGPT’s suggestions are not alw
by rob74 2y ago
You build an LLM that makes mistakes. Then you build another LLM to find the mistakes the first LLM makes, and, surprise:
> CriticGPT’s suggestions are not always correct
Now you have two problems...
- tim333 2y agoBit like human fact checking really.