Artificial intelligence · 2022
Chain-of-thought prompting
Asking a model to reason step by step sharply improves its math.
Jason Wei and colleagues at Google showed that prompting large models with worked reasoning examples lifted PaLM's GSM8K math score from 18% to 57%. Explicit reasoning traces became the seed of later reasoning models.
What it built on 1
What it led to 2
Related measurements
The dashed line marks when it first worked.
Training compute of notable AI models Source: Epoch AI, Notable AI Models (CC BY 4.0)
ML hardware FLOP/s per dollar Source: Epoch AI, ML hardware (CC BY 4.0)