Skip to content
STIMSMITH

Supervised Learning for Coverage-Directed Test Selection in Simulation-Based Verification

Paper WIKI v1 · 5/30/2026

A 2022 arXiv paper that introduces coverage-directed test selection, a supervised-learning method that uses coverage feedback to prioritize randomly generated verification tests likely to increase functional coverage, with the goal of reducing manual constraint writing, resource use, and time to coverage closure.

Overview

Supervised Learning for Coverage-Directed Test Selection in Simulation-Based Verification is a paper submitted to arXiv on 17 May 2022 and last revised as version 3 on 16 October 2022. The arXiv record lists Nyasha Masamba and two other authors. The paper addresses test selection for simulation-based hardware verification, where constrained random test generation is widely used to create verification stimuli. [Submission history and authorship; Problem context]

Problem addressed

The paper observes that randomness in constrained random test generation helps produce diverse tests, but many tests repeatedly exercise the same design logic. In typical verification flows, constraints are often written manually to bias random tests toward interesting, hard-to-reach, or not-yet-tested logic. As verification progresses, however, many constrained random tests have little or no effect on functional coverage. [Problem context]

Proposed method

The paper introduces coverage-directed test selection, described as a method for automatic constraint extraction and test selection. The approach is based on supervised learning from coverage feedback. Instead of simulating every randomly generated test, the method assumes that stimulus generation is significantly cheaper than simulation, generates a large number of tests, and selects a more effective subset for simulation. [Method introduced]

The selection process is biased toward tests estimated to have a high probability of increasing functional coverage, and those tests are prioritized for simulation. [Selection objective]

Reported goals and outcomes

According to the abstract, the method is intended to reduce manual constraint writing, prioritize effective tests, reduce verification resource consumption, and accelerate coverage closure. The paper reports evaluation on a large, real-life industrial hardware design. [Reported outcomes]

Metadata

  • arXiv URL: https://arxiv.org/abs/2205.08524v3
  • Initial submission: 17 May 2022
  • Latest listed revision: version 3, 16 October 2022
  • arXiv categories: cs.AR, cs.AI, cs.LG, cs.SE

LINKED ENTITIES

7 links
Nyasha Masamba AUTHORED_BY
Coverage-Directed Test Selection INTRODUCES
Supervised Learning USES
Simulation-Based Verification USES
constrained-random test generation USES
Coverage Closure EVALUATES
Stimulus Generation MENTIONS

CITATIONS

5 sources
5 citations
[1] Submission history and authorship Supervised Learning for Coverage-Directed Test Selection in Simulation-Based Verification
[2] Problem context Supervised Learning for Coverage-Directed Test Selection in Simulation-Based Verification
[3] Method introduced Supervised Learning for Coverage-Directed Test Selection in Simulation-Based Verification
[4] Selection objective Supervised Learning for Coverage-Directed Test Selection in Simulation-Based Verification
[5] Reported outcomes Supervised Learning for Coverage-Directed Test Selection in Simulation-Based Verification