US 12,306,743 B2
Test case generation
Vijay Arya, Gurgaon (IN); Diptikalyan Saha, Bangalore (IN); Devika Sondhi, Noida (IN); and Kahini Wadhawan, Ferozepur (IN)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Feb. 28, 2023, as Appl. No. 18/115,279.
Prior Publication US 2024/0289260 A1, Aug. 29, 2024
Int. Cl. G06F 11/36 (2025.01); G06F 11/3604 (2025.01); G06F 11/3668 (2025.01); G06N 3/08 (2023.01)
CPC G06F 11/3684 (2013.01) [G06F 11/3608 (2013.01); G06N 3/08 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving at least input data to a computer code and output data that the computer code produces corresponding to the input data;
based on the input data and the output data, building a decision tree that links input conditions to the output data;
using the input conditions, running a condition generator model created by fine-tuning a pre-trained programming language model, the condition generator model predicting a next condition that is likely to occur in the computer code;
based on the next condition, generating input instances and running the computer code using the generated input instances; and
based on output instances output by the computer code based on the generated input instances, detecting at least one flow path in the computer code, which had not been previously built in the decision tree,
wherein fine-tuning the pre-trained programming language model includes:
extracting input variable conditions from a source code sample;
using the extracted input variable conditions, creating a labeled dataset including sequences of conditions; and
fine-tuning at least a last layer of the pre-trained programming language model using the labeled dataset including the sequences of conditions to create the condition generator model.