Subitizing-Inspired_Large_Language_Models_for_Floorplanning
Shao-Chien Lu, Chen-Chen Yeh, Hui-Lin Cho, Yu-Cheng Lin, Rung-Bin Lin
TL;DR
This work tackles VLSI floorplanning by fine-tuning LLMs to predict optimal slicing-floorplans using a post-order tree encoding of recursively generated layouts. A two-stage pipeline combines data generation via recursive slicing with supervised fine-tuning, followed by inference that may require integration with traditional floorplanners to ensure legality and practicality. Empirical results show GPT4o-mini achieving high success and notably low dead space at 16-module configurations, with considerable but diminishing gains as module counts rise to 24. The study demonstrates the potential of LLMs as a complementary tool to conventional floorplanning methods, while highlighting scalability and constraint-compliance challenges for wider real-world adoption.
Abstract
We present a novel approach to solving the floorplanning problem by leveraging fine-tuned Large Language Models (LLMs). Inspired by subitizing--the human ability to instantly and accurately count small numbers of items at a glance--we hypothesize that LLMs can similarly address floorplanning challenges swiftly and accurately. We propose an efficient representation of the floorplanning problem and introduce a method for generating high-quality datasets tailored for model fine-tuning. We fine-tune LLMs on datasets with a specified number of modules to test whether LLMs can emulate the human ability to quickly count and arrange items. Our experimental results demonstrate that fine-tuned LLMs, particularly GPT4o-mini, achieve high success and optimal rates while attaining relatively low average dead space. These findings underscore the potential of LLMs as promising solutions for complex optimization tasks in VLSI design.