LLM-Augmented Bayesian Optimization with Adaptive Policies for Chiplet Design Space Exploration

Name: LLM-Augmented Bayesian Optimization with Adaptive Policies for Chiplet Design Space Exploration
Uploaded: 2026-10-18T23:32:48.039489+00:00
Duration: 334 s
Description: The framework integrates a Large Language Model (LLM) with Bayesian Optimization to automate and improve architectural-level simulation for 2.5D/3D CPU-GPU chiplets. It uses adaptive policy reasoning guided by the LLM to generate search strategies, which then informs a policy-guided Bayesian optimization method, significantly increasing design space exploration efficiency.

The framework integrates a Large Language Model (LLM) with Bayesian Optimization to automate and improve architectural-level simulation for 2.5D/3D CPU-GPU chiplets. It uses adaptive policy reasoning guided by the LLM to generate search strategies, which then informs a policy-guided Bayesian optimization method, significantly increasing design space exploration efficiency.

Open the full episode