Searching protocol for "final-answer"
Format final output as markdown and copy.
Independent solver agents for robust debugging.
Multi-model debates for reliable answers.
AI deliberation for consensus.
Coordinate expert agents for deep reasoning.
Enable reasoning + actions with external tools.
Encourage step-by-step reasoning with CoT
Iterative reasoning with pseudo-code.
Reveal the secret code from asset and script.
Master tool-using agent loops with ReAct.
Long-context reasoning with recursive LLMs.
Advanced reasoning strategies for Gemini prompts.