Part 7/8:
In examining different task requirements, various optimized prompt templates were utilized. For mathematics benchmarks, a clear chain of reasoning was developed, supported by reflective prompts to assess correctness and encourage iterative improvements. Similarly, for coding tasks, prompts were refined to demand explicit correctness confirmations, further enhancing reliability and accuracy.