Albedo tackles model exploit detection issues

Albedo implemented model deduplication and injection detection, but discovered critical bugs: duplicate-submission logic incorrectly flags models as winners, and prompt injection exploits are bypassing validation. Team acknowledged the issues and plans fixes, though exploit models continue scoring near-perfect despite claims they shouldn't be competitive without cheating.

• •Model deduplication and injection detection deployed; re-analyzing incorrectly discarded models
• •Bug found: duplicate-submission code awards points instead of rejecting duplicates
• •Prompt injection exploits scoring ~100% despite validation checks; team investigating
• •GPU failure fixed; new miner validation checks under testing