Closed kgilpin closed 2 months ago
The test generator can get confused about which behavior it's supposed to assert - the bug, or the fix.
So we introduce an extra step that gives it a very direct hint about which way to assert.
On comparison of runs 136 vs 271, there's no difference in the success rate of test patch generation.
The test generator can get confused about which behavior it's supposed to assert - the bug, or the fix.
So we introduce an extra step that gives it a very direct hint about which way to assert.