4# Interview Questions on Large Language Models (LLMs)

❓ Question — What are some of the strategies to 𝗿𝗲𝗱𝘂𝗰𝗲 𝗵𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗶𝗼𝗻 in large language models (LLMs)?

What are some of the strategies to reduce hallucination in large language models (llms)?

Answer:

Hallucinations can be detected in LLM at different levels:
1️⃣ Prompt Level
2️⃣ Model Level
3️⃣ Self-check

A new course launched for interview preparation

We have launched a new course “Interview Questions and Answers on Large Language Models (LLMs)” series.

This program is designed to bridge the job gap in the global AI industry. It includes 100+ questions and answers from top companies like FAANG and Fortune 500 & 100+ self-assessment questions.

The course offers regular updates, self-assessment questions, community support, and a comprehensive curriculum covering everything from Prompt Engineering and basics of LLM to Supervised Fine-Tuning (SFT) LLM, Deployment, Hallucination, Evaluation, and Agents etc.

Detailed curriculum (Get 50% off using coupon code MED50 for first 10 users)

Free self assessment on LLM (30 MCQs in 30 mins)

Prompt Level:

1. Include instructions to the model: Include instruction to the model not to make up stuff on its own. For e.g. “Do not make up stuff outside of given context”
2. Repeat : Repeat most important instructions multiple times.
3. Position: Position most important instruction at the end, making use of latency effect.
4. Parameter: Keep temperature to 0.
5. Restrict: Restrict output to a confined list instead of free float text.
6. Add CoT type instructions : For GPT models “Lets think step by step” works for reasoning task, For PaLM — “Take a deep breath and work on this problem step by step” outperforms.
7. Use Few shot examples — Use domain/use case specific few shot examples, also there is a recent study on new technique called “𝗔𝗻𝗮𝗹𝗼𝗴𝗶𝗰𝗮𝗹 𝗣𝗿𝗼𝗺𝗽𝘁𝗶𝗻𝗴” where model generates its own examples internally which out performs few shot prompting.

𝗔𝗻𝗮𝗹𝗼𝗴𝗶𝗰𝗮𝗹 𝗣𝗿𝗼𝗺𝗽𝘁𝗶𝗻𝗴

8. In Context Learning: Use In-context learning to provide better context to the model.

9. Self-consistency/Voting: Generating multiple answers from the model and selecting most frequent answers.

Model Level:

1. DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models: Simple decoding strategy in large pre-trained LLMs to reduce hallucinations.

GitHub - voidism/DoLa: Official implementation for the paper "DoLa: Decoding by Contrasting Layers…

2. Fine-Tuned model on good quality data — Fine-tuning small LLM model on good quality data has shown promising results as well as help reduce hallucinations.

Self-check:

Methods like chain of verification can help reduce the hallucinations to a great extent, read more here

Chain of verification to reduce hallucinations

Start your interview journey from Question 1:

1# Interview Questions on Large Language models (LLMs

2# Interview Questions on Large Language Models (LLMs)

3# Interview Questions on Large Language Models (LLMs)

Your feedback as comments and claps encourages us to create better content for the community.

Can you give multiple claps? Yes you can