Two failure modes ReAct fixes

Consider the two extremes before ReAct.

Pure reasoning (chain of thought alone). The model thinks beautifully but is cut off from the world. If it does not know a fact, it may confidently invent one. This making-things-up is called hallucination, and it happens because the model has no way to check itself against reality.

Pure acting (tools without reasoning). Imagine a system that can click buttons and run searches but cannot think about why. It tends to flail, taking actions without a coherent plan, easily knocked off track.

ReAct's insight is that reasoning and acting fix each other's weaknesses. Reasoning decides what action to take next and what to make of the result; acting feeds real information back to keep the reasoning honest.