Prompt Engineering about llms4ol HOT 2 CLOSED

Prompt Engineerings:

• Few-shot Prompting: adding examples in prompts to steer the model toward better performances. It enables in-context learning.
• Chain-of-Thought Prompting: Giving instructions to the model for the reasoning task
• Zero-Shot Chain-of-Thought Prompting: It involves adding "Let's think step by step" to the original prompt
• Self-Consistency: Sample multiple and diverse reasoning paths through few-shot CoT and use generation to select the most consistent answers.
• Generate Knowledge Prompting: first generate knowledge from prompt templates, second, generated knowledge used for knowledge-augmented questions to get answers, and a high-confidence response is selected as the final answer!
• Program-aided Language Model‌(PAL): uses LLM to read problems and generate programs as the intermediate reasoning step.
• ReAct: using LLMs to generate both reasoning trace and task-specific actions in an interleaved manner.

from llms4ol.

samples from datasets for analysis:

• seat: he booked their seats in advance
• show: She shows her dogs frequently
• warm: a warm body

T1:

Identify the part of speech tag for the following word.

word: 
part of speech tag: 

from llms4ol.