Leads for review, not findings. Antares-1B scores 0.135 File F1 on its own benchmark: most runs miss the vulnerable file, and a ranked path is not evidence that code is vulnerable — or that unlisted code is safe. Results vary between identical runs.
The model's commands and their output appear here as it works.
This trace is how you check a finding — read it, don't just trust the list.
Nothing yet.