Abstract:Generative Artificial Intelligence (GenAI) has demonstrated its capabilities in the present world that reduce human effort significantly. It utilizes deep learning techniques to create original and realistic content in terms of text, images, code, music, and video. Researchers have also shown the capabilities of modern Large Language Models (LLMs) used by GenAI models that can be used to aid hardware development. Formal verification is a mathematical-based proof method used to exhaustively verify the correctness of a design. In this paper, we demonstrate how GenAI can be used in induction-based formal verification to increase the verification throughput.

What problem does this paper attempt to address?

The paper primarily explores how to utilize Generative Artificial Intelligence (GenAI) to assist in the formal verification process based on induction, aiming to improve verification efficiency and reduce human involvement. Specifically, the paper attempts to address the following issues: 1. **Simplifying the formal verification process**: Formal verification is a mathematical proof method to ensure the functional correctness of hardware designs, but this process is often very time-consuming and requires a significant amount of human effort. Therefore, one of the goals of the paper is to explore how to automatically generate helper assertions through GenAI to simplify this complex process. 2. **Addressing the failure of the induction step**: In induction-based formal verification, if the induction step fails, it usually requires manual analysis of the Counter Example (CEX) and the writing of helper assertions to correct the errors. The paper proposes a method that uses GenAI to automatically generate these helper assertions based on the given design code and counterexamples to address the failure of the induction step. Through the above methods, the paper aims to demonstrate the application potential of GenAI in the field of hardware formal verification and to improve the efficiency and accuracy of formal verification. Additionally, the paper mentions some experimental results that show the effectiveness of this method in practical applications and discusses some limitations and considerations when using GenAI.

Generative AI Augmented Induction-based Formal Verification

Enhancing Reality: Exploring the Potential of Generative Artificial Intelligence

Generative AI Assertions in UVM-Based System Verilog Functional Verification

All Artificial, Less Intelligence: GenAI through the Lens of Formal Verification

Generative Artificial Intelligence: A Systematic Review and Applications

VeCoGen: Automating Generation of Formally Verified C Code with Large Language Models

VerifAI: Verified Generative AI

Investigating the Prospects of Generative Artificial Intelligence

Generative AI Systems: A Systems-based Perspective on Generative AI

Opportunities of using artificial intelligence in hardware verification

AI Horizons: Unveiling the Future of Generative Intelligence

FTA generation using GenAI with an Autonomy sensor Usecase

Deep Verifier Networks: Verification of Deep Discriminative Models with Deep Generative Models

Pair Programming With Generative AI

Neurosymbolic AI for Enhancing Instructability in Generative AI

Generative AI for Visualization: Opportunities and Challenges

Generative AI: A systematic review using topic modelling techniques

Generative artificial intelligence and adversarial network for fraud detections in current evolutional systems

Enhancing Efficiency and Innovation with Generative AI

Generative Artificial Intelligence and GPT using Deep Learning: A Comprehensive Vision, Applications Trends and Challenges

Generative Verifiers: Reward Modeling as Next-Token Prediction