LFX Mentorship Active 2026
AI-assisted extraction of architectural parameters from RISC-V specifications - Part II
RISC-V1 Program Years
AI-assisted extraction of architectural parameters from RISC-V specifications - Part II
Open Source
About the Organization
<p>This mentorship proposal extends an effort already underway to extract architectural parameters from the RISC-V ISA Manual using AI. It originally started as an RVI Mentorship for Spring 2026, and continued under the Parameter SIG. Although the results have improved steadily, we need further improvement in quality and implementation robustness. Hence the Fall mentorship proposes to:</p><p><br></p><p>1) Continue trying to find architectural parameters with LLMs in the RISC-V privileged and unprivileged spec, extending the initial work done in Spring 2026 mentorship. Use as training examples subsets of parameters from the manually created lists (and try to recreate the full lists) from the following efforts:</p><p><br></p><p>a. ISA Manual (per-chapter params in yaml format)</p><p>b. Google Drive (“keyword_matches” spreadsheet)</p><p>c. UDB yaml (enhance the current parameter recall process).</p><p><br></p><p>2) Extend the current classification scheme for parameters as needed.</p><p><br></p><p>3) Develop AI coding agents and skills for reproducible runs and reusable workflows (extending the prompt and context management in the current flow).</p><p><br></p><p>4) Explore integration of the tools developed for the ISA Manual based flow (in 1.a) to export the parameters in UDB yaml format.</p><p><br></p><p>5) Create a Github PR to publish the final reviewed parameter files in the appropriate repo and follow up with the maintainers on merging them.</p><p> </p>
Primary Technologies & Languages
generative aiisa specificationsparameterized modeling
Technical Topics & Focus Areas
generative aiisa specificationsparameterized modeling
Participation Years
Completed Projects Distribution
Annual Completed Projects History
AI-assisted extraction of architectural parameters from RISC-V specifications - Part IIHover over bars to view completed counts
4
3
2
1
0
2026
Projects1
Years Active1
Stackgenerative ai
TierTier 1
Past Projects
Showing 1 of 1 projects. Click any project card for scope, mentors, and proposal studio.
AI-assisted extraction of architectural parameters from RISC-V specifications - Part II Mentorship Project
Intermediate<p>This mentorship proposal extends an effort already underway to extract architectural parameters from the RISC-V ISA Manual using AI. It originally started as an RVI Mentorship for Spring 2026, and continued under the Parameter SIG. Although the results have improved steadily, we need further improvement in quality and implementation robustness. Hence the Fall mentorship proposes to:</p><p><br></p><p>1) Continue trying to find architectural parameters with LLMs in the RISC-V privileged and unprivileged spec, extending the initial work done in Spring 2026 mentorship. Use as training examples subsets of parameters from the manually created lists (and try to recreate the full lists) from the following efforts:</p><p><br></p><p>a. ISA Manual (per-chapter params in yaml format)</p><p>b. Google Drive (“keyword_matches” spreadsheet)</p><p>c. UDB yaml (enhance the current parameter recall process).</p><p><br></p><p>2) Extend the current classification scheme for parameters as needed.</p><p><br></p><p>3) Develop AI coding agents and skills for reproducible runs and reusable workflows (extending the prompt and context management in the current flow).</p><p><br></p><p>4) Explore integration of the tools developed for the ISA Manual based flow (in 1.a) to export the parameters in UDB yaml format.</p><p><br></p><p>5) Create a Github PR to publish the final reviewed parameter files in the appropriate repo and follow up with the maintainers on merging them.</p><p> </p>