MALLA REDDYENGINEERINGCOLLEGEFORWOMEN
(Autonomous Institution–UGC Govt. of India)
Accredited by NAAC with A+ grade
Affiliated to JNTUH, Approved by AICTE, ISO 9001-2015 Certified
COURSE: SUBJECT: Large Language
DEPARTMENT: CSE
IV B. TECH I SEM Models (LLM)
UNIT–I R22 REGULATION ACADEMIC YEAR:2025-2026
TUTORIALSHEET-I
SET-I
[Link] QUESTION BTL CO
1 Define Natural Language Processing (NLP) and list any three
common tasks it is used for. Remember CO1
2 Explain how transformer-based models differ from recurrent neural
Understand
networks (RNNs) in processing language data. CO1
3 Given a sample sentence, demonstrate how a language model could
Apply CO1
be used to generate a summary of it.
4 Compare the advantages and limitations of RNNs and Transformer
models for real-time applications such as chatbots. Analyze CO1
5 Design a conceptual solution using a hybrid language model to
improve the accuracy of a machine translation system. Create CO1
Signature of Faculty Signature of HOD
.
MALLA REDDYENGINEERINGCOLLEGEFORWOMEN
(Autonomous Institution–UGC Govt. of India)
Accredited by NAAC with A+ grade
Affiliated to JNTUH, Approved by AICTE, ISO 9001-2015 Certified
COURSE: SUBJECT: Large Language
DEPARTMENT:CSE
IV B. TECH I SEM Models (LLM)
UNIT–II R22 REGULATION ACADEMIC YEAR:2025-2026
TUTORIALSHEET-2
SET-II
[Link] QUESTION BTL CO
1 What is the purpose of the self-attention mechanism in transformer
architectures? Remember CO1
2 Describe the key differences between BERT and RoBERTa in terms
of their training methodology CO2
Understand
3 Given a text classification dataset, explain how you would fine-tune
Apply CO2
a pre-trained BERT model for this task
4 Analyze the limitations of transformer-based models when applied to
long text documents and suggest potential solutions Analyze CO2
5 Design a custom NLP pipeline using RoBERTa for sentiment
analysis in multilingual social media data. Include preprocessing,
Create CO2
model adaptation, and evaluation strategies.
Signature of Faculty Signature of HOD
MALLA REDDYENGINEERINGCOLLEGEFORWOMEN
(Autonomous Institution–UGC Govt. of India)
Accredited by NAAC with A+ grade
Affiliated to JNTUH, Approved by AICTE, ISO 9001-2015 Certified
COURSE: SUBJECT: Large Language
DEPARTMENT:CSE
IV [Link] I SEM Models (LLM)
UNIT–III R22 REGULATION ACADEMIC YEAR:2025-2026
TUTORIALSHEET-3
SET-III
[Link] QUESTION BTL CO
List three common preprocessing steps applied to large-scale text
1 datasets before training a language model. Remember CO3
Explain why tokenization is critical in preparing data for large
2 language model training. Understand CO3
Given a noisy text dataset, describe the preprocessing pipeline you
3 Apply CO3
would implement before feeding it into a transformer-based model.
Evaluate the effectiveness of perplexity versus accuracy as evaluation
metrics for language model performance. Which is more suitable for
4 Analyze CO3
generative models and why?
Design a training strategy for a domain-specific language model that
addresses issues such as overfitting, long training time, and data
5 Create CO3
imbalance.
Signature of Faculty Signature of HOD
MALLA REDDYENGINEERINGCOLLEGEFORWOMEN
(Autonomous Institution–UGC Govt. of India)
Accredited by NAAC with A+ grade
Affiliated to JNTUH, Approved by AICTE, ISO 9001-2015 Certified
COURSE: SUBJECT: Large Language
DEPARTMENT:CSE
IV [Link] I SEM Models (LLM)
UNIT–IV R22 REGULATION ACADEMIC YEAR:2025-2026
TUTORIALSHEET-4
SET-IV
[Link] QUESTION BTL CO
1 What is a multimodal language model? Provide one example of such a Remember CO4
model.
2 Describe how explainability methods like attention visualization help Understand CO4
interpret transformer-based models.
3 Suppose you are deploying a large language model in a healthcare Apply CO4
chatbot. How would you apply bias detection methods to ensure
ethical use?
4 Suppose you are deploying a large language model in a healthcare chatbot. Analyze CO4
How would you apply bias detection methods to ensure ethical use?
5 Propose a research idea that addresses one current limitation of large Create CO4
language models, such as hallucination or limited world knowledge.
Signature of Faculty Signature of HOD
MALLA REDDYENGINEERINGCOLLEGEFORWOMEN
(Autonomous Institution–UGC Govt. of India)
Accredited by NAAC with A+ grade
Affiliated to JNTUH, Approved by AICTE, ISO 9001-2015 Certified
.
COURSE: SUBJECT: Large Language
DEPARTMENT:CSE
III [Link] I SEM Models (LLM)
UNIT–V R22 REGULATION ACADEMIC YEAR:2025-2026
TUTORIALSHEET-5
SET-V
[Link] QUESTION BTL CO
List three real-world applications of large language models in industry
1 sectors like healthcare, finance, or customer service. Remember CO5
Explain how large language models can be applied to improve speech
2 recognition systems. CO5
Understand
You are tasked with building a text-to-image captioning system.
3 Describe how you would integrate a large language model into your Apply CO5
pipeline.
Analyze a case study where an LLM deployment failed due to poor
4 maintenance or scalability issues. What lessons can be learned? Analyze CO5
Design a complete project proposal that uses a large language model to
solve a specific problem (e.g., legal document summarization or
5 educational tutoring). Include objectives, methodology, and expected CO5
Create
outcomes.
Signature of Faculty Signature of HOD