The output in feedback should correspond with the response in history

The response saved the history is always the first one: https://github.com/ASSERT-KTH/DET-Gen/blob/main/src/test_generator.py#L88

But the output in the feedback may be computed for another response, if the first one is invalid: https://github.com/ASSERT-KTH/DET-Gen/blob/main/src/code_runner.py#L148

Start by comparing the history-saving logic at src/test_generator.py#L88 with the feedback execution path at src/code_runner.py#L148. Trace how responses are selected when the first response is invalid, then verify that the response recorded in history is the one whose output appears in feedback. Done means these responses remain aligned in that case.

Written by the indexing model from the issue text.