Master5401

AI-Assisted RISC-V Parameter Extraction

Applicant: Shakunth Srinivasan LFX Mentorship Spring 2026

Project Overview

This submission addresses the challenge of automating the extraction of architectural parameters from RISC-V specifications. My solution benchmarks two LLMs (GPT-4o and DeepSeek-V3) and implements a specific taxonomy to categorize parameters into UDB-compatible formats.

1. Methodology

Model Selection

• GPT-4o: Selected for its high reasoning capability and adherence to complex formatting instructions.