feat: add failed flag submissions csv as part of dataset

[GangGreenTemperTatum]

add failed flag submissions csv as part of dataset

Key Changes:

• add failed flag submissions csv as part of dataset

Added:

• add failed flag submissions csv as part of dataset

---

Generated Summary:

• Changed the .gitignore entry from “datasets/” to “dataset/” and added an exception for “failed_flag_submissions.csv”, which will now be tracked.
• Added a new file “dataset/failed_flag_submissions.csv” containing numerous flag submission records (including model display names, challenges, failed attempt counts, and last flag submissions).
• The update ensures that the critical flag submission data is not excluded by git, which may affect automated flag processing or tracking.

This summary was generated with ❤️ by rigging

[linear]

ENG-2239 Add failed_flags.csv to AIRTBench-Code repo and update notes

ads said:

FYI i also want to include the failed_flags.csv so throwing that in the AIRTBench-Code repo and added a bit of notes to the paper under the Spurious Elaboration as i think it'd be good to promote additional transparency into reasoning patterns and failure modes