Reproducible backdoor-detection research harness with harmless toy controls, paired-response evaluation, and JSON/HTML reports.

ai-safety backdoor-detection benchmark llm-security open-weight-models reproducibility
3 Open Issues Need Help Last updated: Oct 15, 2026

Open Issues Need Help

View All on GitHub

Reproducible backdoor-detection research harness with harmless toy controls, paired-response evaluation, and JSON/HTML reports.

Python
#ai-safety#backdoor-detection#benchmark#llm-security#open-weight-models#reproducibility

Reproducible backdoor-detection research harness with harmless toy controls, paired-response evaluation, and JSON/HTML reports.

Python
#ai-safety#backdoor-detection#benchmark#llm-security#open-weight-models#reproducibility

Reproducible backdoor-detection research harness with harmless toy controls, paired-response evaluation, and JSON/HTML reports.

Python
#ai-safety#backdoor-detection#benchmark#llm-security#open-weight-models#reproducibility