All Game Theory Machine Learning Artificial Intelligence Algorithms NLP General

Projects

Bounded Rationality driven resource allocation in Cournot Games

May 2025 – Present | Advisor: Prof. Pavan Tallapragada

Multi-agent based model with bounded rationality, focusing on resource allocation in Cournot competition settings.

[GitHub] [Draft]

Game TheoryMachine LearningOpinion Dynamics

STREAM

Summer 2024 | Advisor: Prof. Y Narahari

Simulation and Traffic Routing for Efficient Allocation and Management. Tool Developed to test and analyse the Braess's Paradox behaviour in traffic networks.

[GitHub]

Game TheoryAlgorithmsNetwork Analysis

Lurking in social networks

Winter 2024 | Advisor: Prof. C Pandurangan

Investigating the dynamics of user behavior and information spread in social networks. Designing and identifying lurking behaviour in Social networks to enhance ranking algorithms.

Social Network AnalysisAlgorithms

stratepy

Mar 2024 | Independent Study

Efficient python library designed to model and simulate game theoretical scenarios. Also contains implementation of voting and auction systems.

[GitHub] [PyPI]

Game TheoryAlgorithmsSimulaitonsPython Library

Equilibrium Computation in Multi-agent Systems

Summer 2024 | Independent Study

Fast heuristics for computing equilibria in structured games using convex optimization and randomized techniques.

[GitHub]

AlgorithmsGeneral

Publications

HinglishEval: Evaluating the Effectiveness of Code-Generation Models on Hinglish Prompts

Siddharth Reddy, Mrigank Pawagi, Kintan Saha, Annirudh Gupta

Submitted to ACM COMPUTE 2024

[Springer] [Code]

@InProceedings{10.1007/978-3-031-84391-4_2,
author="Pawagi, Mrigank
and Gupta, Anirudh
and Rolla, Siddharth Reddy
and Saha, Kintan",
editor="Karkare, Amey
and Prasad, Prajish
and Raman, Arun",
title="HinglishEval: Evaluating the Effectiveness of Code-Generation Models on Hinglish Prompts",
booktitle="Computing Education Research",
year="2025",
publisher="Springer Nature Switzerland",
address="Cham",
pages="18--29",
abstract="Code-generation Models are Large Language Models (LLMs) that are fine-tuned to generate code from natural-language prompts. Prior work shows that such models can democratize programming by translating clear English-language prompts into code for simple programming tasks. In this paper, we explore whether this democratization can extend to novice programmers who lack proficiency in English but are able to craft clear prompts in another language. Specifically, we consider prompts in Hinglish, a mixture of Hindi and English that many students in India are comfortable with. We make two contributions. First, we propose a semi-automated technique to translate English prompts into Hinglish, and we use this technique to create HinglishEval: a Hinglish translation of the widely-used code-generation benchmark HumanEval. Second, we compare the performance of several popular open- and closed-source code-generation models on Hinglish and English prompts. Our findings suggest that although code-generation models are generally more effective at generating accurate code from English prompts, their efficacy with Hinglish prompts is promising.",
isbn="978-3-031-84391-4"
}

Machine LearningArtificial IntelligenceNLP