MALLA REDDY ENGINEERING COLLEGE FOR WOMEN
(Autonomous Institution – UGC Govt. of India)
Accredited by NAAC with A+ grade
Affiliated to JNTUH, Approved by AICTE, ISO 9001-2015Certified
COURSE: IV [Link] SUBJECT: LLM
DEPARTMENT:CSE
I SEM
UNIT – I,II,III R22 REGULATION ACADEMIC YEAR:2025-2026
ASSIGNMENT SHEET-I
SET-I
[Link] QUESTION BTL CO
1 Define Natural Language Processing (NLP). Mention any two 1 1
applications of NLP in daily life.
Explain the purpose of the self-attention mechanism in the
2 2 2
transformer architecture.
Given a sample noisy text corpus, describe the steps to
3 3 3
preprocess it for training a language model.
Compare and contrast explainability and interpretability in the
4 4 4
context of large language models.
5 Design a mini project that uses an LLM to automate resume 6 5
screening for recruitment.
SET-II
[Link] QUESTION BTL CO
1 Describe the evolution of language models from RNNs to 2 1
transformer-based architectures.
2 Demonstrate how you would fine-tune a RoBERTa model for 3 2
sentiment analysis.
3 Analyze the limitations of accuracy and BLEU score as 4 3
evaluation metrics for LLMs.
4 Evaluate the ethical concerns that arise from deploying LLMs 5 4
in education systems.
5 6 5
Propose an industry-based solution using an LLM for multilingual
document classification.
SET-III
[Link] QUESTION BTL CO
1 Apply your understanding of LLMs to describe how a chatbot 3 1
would generate responses to user queries.
2 What are BERT and RoBERTa? List one key difference between 1 2
them.
3 Explain the role of learning rate and batch size in training 2 3
large-scale language models.
4 Design a strategy for integrating vision and language in a multimodal 6 4
language model.
5 Analyze a real-world case study where LLM deployment failed 4 5
or led to unexpected outcomes
SET-IV
[Link] QUESTION BTL CO
1 Analyze how transformer-based models improve upon RNNs in 4 1
terms of performance and scalability.
2 Critically evaluate the advantages and disadvantages of 5 2
transformer models in NLP tasks.
3 Design a training pipeline for domain-specific legal document 6 3
processing using LLMs.
4 What are multimodal models? Give an example where text and 1 4
image inputs are used..
5 Describe the best practices in maintaining LLMs after 2 5
deploying them in production.
SET-V
[Link] QUESTION BTL CO
1 Propose a new hybrid model combining RNN and transformer 6 1
components for low-resource NLP tasks.
2 Analyze how pretraining objectives in BERT affect its 4 2
performance in downstream tasks.
3 What is overfitting in the context of LLM training? How can it 2 3
be prevented?
4 Explain how attention heatmaps can be used for interpreting 1 4
transformer predictions.
5 Apply a large language model to automate customer support 3 5
ticket categorization.
SET-VI
[Link] QUESTION BTL CO
1 Evaluate the suitability of LLMs for tasks like text 6 1
summarization compared to traditional rule-based systems.
2 Design a fine-tuning strategy using BERT for a question- 5 2
answering system in the education domain.
3 Explain the significance of using diverse datasets during the 1 3
training of LLMs.
4 Apply ethical evaluation frameworks to assess bias in a 3 4
language model trained on internet data.
5 List three real-world applications of LLMs in speech, vision, 1 5
and multimodal systems.
Signature of Faculty Signature of HOD
