3 ms·
> Bigger models hallucinate less. I'm skeptical. Based on what research?
by xpe 3y ago
> Bigger models hallucinate less.
I'm skeptical. Based on what research?
- famouswaffles 3y agoGPT-4 hallucinates a lot less than 3.5. Same with the Claude Models. This is from personal experience. There are also benchmarks (like TruthfulQA) that try to measure hallucinations that show the same thing.