Skip to content
BC 2026 release wave 2unofficial launch event map, by waldo

Features › Developer tools

Contamination detection

GA · 3 min in 1 video · dev relevance medium · not in the documented features baseline (docs checked 2026-10-02)

Contamination detection

Models get only the repo name and bug description and must name files to modify. Only one task matched across 101 tasks and three models.

Area: Developer tools. Status: GA. Developer relevance: medium. Airtime: 3 min across 1 video, demoed.

Status evidence

  • Nothing said about status; launch event convention: generally available unless stated otherwise.

Where they talk about it

Quotes

  • 7:54 "for all three models they only managed to uh come up with one out of 101 tasks" - Key number showing very little sign of memorization of the data set.

Caveats mentioned

  • The one match was a task whose description contained the exact file path
  • Monitoring will be continuous, not a one-time effort

Documented features match

  • No documented item matched. BC-Bench not documented

Docs checked on 2026-10-02: the match is against Microsoft's what's new pages for the wave as they were on that date. Microsoft keeps filling the documentation, so "no documented item" can go stale. The wider product documentation on learn.microsoft.com was searched for this feature on the same date; the feature's data lists an article when one was found.

Tags: testing, agents, al

Generated by pipeline step 06. Source: YouTube auto-captions of the launch event videos, see the deep links. Not official.