Your question is ReAct vs CoT vs ToT. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
You are choosing a prompting strategy for an LLM feature that must handle complex, multi-step tasks. Some tasks are pure reasoning, while others require calling tools, checking intermediate results, or exploring multiple possible paths before answering.
Explain the trade-offs between ReAct, Chain-of-Thought (CoT), and Tree-of-Thoughts (ToT) prompting strategies for complex, multi-step tasks. When would you prefer each one, and how do their differences affect reliability, cost, latency, and hallucination risk?