2 ms·
Bigger models hallucinate less. and we don't call it hallucinations but gofai mispredicts plenty.
by famouswaffles 3y ago
Bigger models hallucinate less.
and we don't call it hallucinations but gofai mispredicts plenty.
- 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.