paraacha @talhaparacha.com · Dec 16

We also experiment with the Temperature parameter to control sampling strategy. We find that sampling in a conservative way makes more instances valid, at the expense of diversity in features. And that the other extreme adds too much randomness, which also hurts testing. 9/n

0 likes 1 replies

?

Replies

paraacha · Dec 16

For full details, and to understand the security implications of our findings, please find a preprint here (talhaparacha.com/icse_preprin...). We’ll be at #icse2026 in April to discuss results in person. 10/n